The Technology Modernization Fund is asking federal agencies to submit permitting-technology and artificial-intelligence proposals by July 24, 2026, as it races to commit a remaining $200 million before its authority to make new investments expires on September 30 unless Congress intervenes. The call is less a routine funding round than a compressed test of whether government can finance urgent modernization at the speed its own technology policy now demands. Agencies have an opportunity to bypass some of the delay built into normal budget and procurement cycles, but they must arrive with projects mature enough to survive rapid scrutiny. The larger contradiction is stark: Washington wants faster permitting and responsible AI adoption while allowing one of its principal modernization mechanisms to approach a statutory cliff.
As first reported by FedScoop, Jessie Posilkin, acting director of the Technology Modernization Fund, is being unusually explicit about both the targets and the urgency of the current proposal round. The fund wants projects that can speed and scale permitting work supported by the Council on Environmental Quality, as well as projects that can advance USAi adoption in agencies where capability or infrastructure gaps are blocking responsible use.
Those priorities are broad enough to attract substantial interest but narrow enough to reveal the administration’s immediate technology agenda. Permitting modernization concerns the machinery through which government approves consequential physical projects; AI adoption concerns the machinery through which government increasingly expects its employees to analyze, write, code, and make decisions.
Both are bottleneck problems. Permitting can become trapped in fragmented systems, inconsistent data, manual handoffs, and weak coordination between agencies. AI can become trapped at the pilot stage because agencies lack prepared data, secure infrastructure, governance controls, acquisition pathways, or the capacity to operate tools beyond a demonstration.
The TMF’s official call says proposals outside the two priority areas remain welcome. In practice, however, agencies now know where the strongest policy signal lies, and they know they are competing against a calendar that leaves little space for vague concepts, unfinished business cases, or extended internal negotiation.
“TMF was made for moments like this, where agencies can’t afford to wait on budget and procurement cycles,” Posilkin wrote in announcing the push. That is an argument for the fund’s design, but it is also an indictment of the system around it: if normal budgeting cannot respond quickly enough to clearly identified technology needs, then the government’s emergency lane becomes indispensable.
The trouble is that the emergency lane is itself running out of road.
Yet the operational needs behind them overlap. Both require agencies to structure and share data, replace fragile manual workflows, build secure interfaces, coordinate across organizational boundaries, and define who is accountable when technology changes the way decisions are made.
The official TMF material describes the permitting track as a way to expand implementation scope, improve interagency data sharing and coordination, or accelerate high-impact work already underway. That framing favors projects with an established policy foundation and a plausible path to execution, rather than speculative efforts that begin with a technology and search for a government problem to attach to it.
The AI call is similarly oriented toward implementation rather than enthusiasm. It seeks high-impact, shovel-ready projects that can prepare data and infrastructure, pilot emerging tools while refining guardrails, or produce secure capabilities at enterprise scale.
That is a meaningful distinction. Federal agencies have no shortage of AI demonstrations, vendor briefings, experimentation groups, and employee interest; what many lack is the institutional substrate needed to turn a promising model into a reliable service.
The gap can include identity and access management, logging, data classification, records handling, privacy review, testing, procurement, workforce skills, integration with existing systems, and mechanisms for monitoring outputs after deployment. An AI pilot can be produced by a small team with limited data, but responsible agency-wide adoption requires controls that survive changes in personnel, vendors, models, and policy.
The TMF is therefore not simply offering to purchase AI tools. It is positioning itself to finance the less glamorous foundations without which those tools remain isolated experiments—or become unmanaged liabilities.
That observation cuts against the simplistic idea that AI automatically reduces the cost and complexity of government technology. Generative tools may help employees create code, documentation, queries, workflows, or analyses more quickly, but greater production speed can also multiply the number of artifacts an agency must review, secure, maintain, and eventually retire.
AI-assisted coding, for example, does not remove the need for software inventories, source control, dependency management, access restrictions, testing, security review, documentation, and ownership. It can instead produce more code across more teams, sometimes written by employees who could not have created the same software unaided and may be less prepared to recognize subtle defects.
The resulting challenge is not that AI necessarily generates bad work. It is that it changes the ratio between creation and oversight: agencies may suddenly be able to produce applications, scripts, analyses, and automations faster than their existing governance systems can classify and supervise them.
The same issue applies to information management. An employee who uses an AI tool to summarize documents, generate correspondence, analyze case files, or build an internal assistant may create new copies, derived records, prompts, logs, embeddings, indexes, and outputs. Each can carry retention, privacy, security, evidentiary, or public-disclosure implications.
Responsible adoption consequently requires more than a list of approved models. Agencies need enforceable rules about what information can enter a system, how outputs may be used, what activity must be logged, when human review is required, how results are challenged, and who has authority to shut down a deployment that behaves unexpectedly.
This is why the TMF’s emphasis on agencies with capability and infrastructure gaps matters. A weak foundation cannot be repaired merely by placing an AI interface over it. In some cases, AI will expose and amplify weaknesses that were already present in an agency’s data architecture, access controls, documentation, or software-development practices.
The best proposals will recognize that reality. They will describe AI as a governed service operating inside an accountable technical environment—not as a magic layer that compensates for everything underneath it.
A polished public portal can still conceal manual re-entry, emailed spreadsheets, inconsistent identifiers, and disconnected status systems behind the scenes. Applicants may receive a better-looking front end while government employees continue reconciling records by hand.
The TMF call’s emphasis on sharing data and coordinating with other agencies points toward the harder work. Effective modernization may require common data definitions, interoperable interfaces, reliable status exchange, ownership agreements, and processes that allow one agency to trust information supplied by another.
Technology can accelerate a review only when the underlying policy and operational handoffs are sufficiently clear to encode. If two agencies disagree over responsibility, required evidence, or the meaning of a status field, a new system may digitize that dispute rather than resolve it.
The strongest permitting proposals should therefore connect technical deliverables to measurable changes in the process. Faster document intake matters, but so do reduced duplicate submissions, fewer manual transfers, clearer status visibility, earlier identification of incomplete applications, and better coordination between reviewing bodies.
This is also where the TMF model can have value beyond a conventional appropriation. A modernization fund can support work that crosses organizational lines and does not fit cleanly into one office’s existing operations budget, particularly when the benefit is distributed among several agencies or accrues over multiple years.
But cross-agency projects are difficult to assemble quickly. They require sponsorship, agreements, architecture decisions, and clarity about who will operate the resulting service after the initial investment. The July 24 deadline leaves little time to create that alignment from scratch.
At government-wide scale, however, it is a constrained portfolio. It must be weighed against aging systems, cybersecurity requirements, citizen-service failures, data remediation, cloud migration, interoperability needs, and now an expanding set of AI infrastructure and governance demands.
“We are a country that continues to have really significant service delivery challenges, which is what the government should ultimately be providing,” Posilkin said. “And when you have it set up that way, $200 million is actually quite the drop in the bucket.”
Her point is not simply that government technology is expensive. It is that the cost of modernization is distributed unevenly and often becomes visible only when a legacy system fails, a security incident occurs, or a high-profile service cannot meet demand.
A fund of this size also cannot rescue every deserving project. Selection becomes an exercise in leverage: the board must identify investments that can unlock larger improvements, create reusable components, remove a particularly consequential bottleneck, or provide evidence that supports broader adoption.
That raises the cost of choosing poorly. A project that begins with an unclear problem, depends on unresolved policy decisions, or cannot survive beyond the initial funding period consumes capital that cannot be recovered merely by producing a demonstration.
The reference to 100% of agencies repaying TMF loans is therefore important but not sufficient. Repayment can sustain the revolving model, yet it does not eliminate the timing mismatch between repayments arriving and new needs emerging. Nor does it make the fund large enough to match the growing scale and scope of government technology challenges.
Repayment requirements can also influence which projects agencies bring forward. Projects with obvious savings or measurable cost avoidance may be easier to justify than public-service improvements whose benefits are substantial but difficult to convert into a repayment schedule.
This tension has followed modernization funding for years: government wants projects to demonstrate financial discipline, but some of its most important outcomes—less citizen frustration, more reliable access, shorter processing times, better security, and reduced operational risk—do not appear as simple revenue.
For the current call, agencies will need to connect mission value to a credible investment structure. A proposal cannot rely solely on the importance of AI or permitting as a policy area; it must explain why TMF financing is necessary, why the work is ready now, and how the investment produces durable value.
But the bill has not received a floor vote in either chamber. Its treatment of the fund may change, and inclusion in committee legislation should not be mistaken for final reauthorization or a stable long-term mandate.
Even if enacted at the stated amount, $5 million would be minor next to the $200 million Posilkin said remained last month. It would signal that Congress had not entirely abandoned the mechanism, but it would not provide the scale implied by the administration’s ambitions for AI, permitting, cybersecurity, and citizen services.
The more immediate issue is authorization. The fund’s authority to make new investments ends on September 30 absent congressional action, leaving TMF staff to evaluate proposals, conduct due diligence, negotiate terms, and complete selections within a hard and rapidly approaching boundary.
That is why the July 24 submission date cannot be treated like an ordinary grant deadline. Waiting until the final day may technically preserve eligibility, but it reduces the opportunity to answer questions, resolve weaknesses, or adjust a proposal before the fund’s own decision window closes.
“This call is moving fast, partly out of necessity,” Posilkin wrote. “We have a very narrow window between now and September 30, when our authorization to make new investments ends (absent congressional action, of course…).”
Congress could extend that window. Until it does, agency leaders have to plan against the authority that exists, not the authority they hope will exist.
That distinction is familiar to federal IT executives. Long-term modernization is frequently expected to proceed through short-term legislative decisions, continuing uncertainty, and annual funding processes that do not align well with multi-year technical delivery.
The TMF was designed in part to escape that trap. Its current predicament shows that an alternative funding mechanism remains dependent on the same political calendar it was meant to help agencies navigate.
July 24, 2026 — Agencies must submit initial project proposals for consideration under the call, with earlier submissions strongly encouraged.
September 30, 2026 — TMF’s authorization to make new investments ends unless Congress acts; AI proposals submitted under the call can be considered for selection and announcement by this date.
Fiscal 2027 — Potential reauthorization remains unresolved, while the current version of the relevant appropriations bill includes $5 million and awaits a floor vote in both chambers.
For an AI proposal, maturity should include a defined use case and the data necessary to support it. It should also include a realistic account of security, privacy, records, accessibility, evaluation, workforce impact, and human oversight.
A proposal that says an agency wants to “use AI to improve efficiency” is not a project plan. It does not identify the decision, workflow, service, dataset, user population, risk, or result that would allow reviewers to judge whether the investment is worthwhile.
A stronger proposal might describe a bounded operational process, explain why current systems or staffing cannot meet the need, and define what success would look like without presuming that model adoption is itself the desired outcome. It would separate the cost of the AI capability from the cost of preparing the agency to operate it responsibly.
Permitting proposals face an equivalent test. A project should identify the stage at which delay occurs, the agencies or parties involved, the information exchanged, and the technical or procedural constraint that prevents improvement.
Claims about acceleration require a baseline. If an agency cannot describe current processing time, manual effort, error rates, resubmissions, backlogs, or coordination delays, it may struggle to demonstrate that a new system produced a meaningful result.
The phrase shovel-ready also implies that projects cannot depend on a long chain of unresolved decisions after selection. TMF reviewers are likely to be wary of proposals requiring years of preliminary analysis before users or the public receive a benefit.
That does not mean agencies should hide complexity. The compressed window makes honest dependency mapping more valuable, because an optimistic schedule built on missing approvals or uncertain data access can fail faster than a cautious one.
The best candidate is not necessarily the project promising the most dramatic transformation. It is the project with a consequential problem, a credible delivery path, manageable dependencies, measurable outcomes, and leadership prepared to make decisions when implementation becomes difficult.
AI makes that operational burden more visible. Models, integrations, policies, and usage patterns can change, requiring ongoing evaluation rather than a single acceptance test at deployment.
An enterprise AI capability may need monitoring for accuracy, drift, inappropriate access, misuse, cost escalation, and changes in the source data or model. If an agency lacks the people and processes to perform that work, the project is not truly ready for enterprise operation.
Permitting platforms have similarly persistent responsibilities. Data standards evolve, participating organizations change, regulations are updated, interfaces fail, user expectations rise, and the number of records grows.
TMF funding can help an agency overcome the initial investment barrier. It cannot substitute indefinitely for a product owner, operating budget, support model, security program, or governance structure.
Applicants should consequently treat sustainability as part of the architecture. They need to identify who owns the capability after the funded milestones, how ongoing costs will be covered, and whether the agency can maintain critical components without permanent dependence on one contractor.
The repayment condition reinforces this need. If an agency must both operate the service and repay the investment, an unrealistic cost model can undermine the project after the visible implementation phase has ended.
A credible application should distinguish among one-time modernization costs, recurring operational costs, repayment obligations, and expenses expected to be absorbed by existing programs. Blurring those categories may make a proposal appear cheaper, but it transfers risk into later budget years.
A flood of late submissions can create pressure to triage quickly, prioritize proposals requiring less clarification, and spend limited staff time on projects that already present a coherent case. Agencies that submit early may give reviewers more opportunity to identify missing information while there is still time to respond.
This does not mean early submission guarantees selection. It means that proposal quality includes the ability to be understood and evaluated within the available window.
Federal technology applications can become dense compilations of policy language, broad benefits, architecture terminology, and optimistic milestones. Under a compressed review schedule, that style becomes a practical disadvantage.
The proposal needs to establish the problem, intervention, outcome, risk, cost, and implementation path without forcing reviewers to reconstruct the logic from attachments. It should be specific enough for technical scrutiny but clear enough for decision-makers to see why the project matters.
Applicants should also anticipate that the TMF team may challenge assumptions about scope, repayment, acquisition, security, and sequencing. A project team that cannot answer promptly because authority is scattered among several offices will lose time it does not have.
The July 24 date is therefore not the true beginning of the process. For serious applicants, work must begin well before submission, and internal approval channels should be prepared to handle follow-up at a pace that normal agency governance may not naturally support.
The harder question is whether agreement on outcomes will produce durable support for the financing mechanism. Congress can endorse modernization in principle while disagreeing over spending levels, executive priorities, oversight, repayment, or the size and role of a central fund.
The $5 million included in the current appropriations bill demonstrates the difference between recognition and scale. It acknowledges that TMF remains part of the government’s technology-financing framework, but it does not resolve how large or durable that role should be.
This matters because agencies make different choices when a program is perceived as temporary. They may rush proposals, narrow ambitions, defer cross-agency work, or avoid investing staff time in applications whose funding mechanism may disappear.
TMF staff face a similar problem. Building expertise in evaluating and overseeing complex modernization projects is itself a long-term institutional investment. Uncertain authority makes it harder to plan the portfolio, retain capability, and provide agencies with predictable guidance.
If lawmakers value the fund, the useful debate is not simply whether to keep its name in an appropriations bill. It is whether TMF should operate as an enduring government-wide investment mechanism with enough authority and capital to support a deliberate portfolio.
Without that certainty, every urgent call risks becoming a race to obligate funds before another deadline. That may produce successful projects, but it is an inefficient way to manage the government’s technology estate.
Permitting teams should resist treating a new portal as the full solution when the actual delays lie in data exchange or organizational handoffs. AI teams should resist treating model access as enterprise adoption when the actual gaps concern governance, infrastructure, and operational ownership.
The TMF can supply capital and specialized oversight. It cannot manufacture consensus inside an agency, create a reliable dataset from undocumented records overnight, or assign accountability where leaders have avoided doing so.
That makes this round a useful test of federal technology maturity. It will show whether agencies have moved beyond broad modernization strategies and accumulated a pipeline of investment-ready projects that can be activated when financing becomes available.
It will also test the fund’s ability to preserve quality under deadline pressure. Moving quickly is valuable only if the resulting investments remain technically credible, governable, and sustainable.
There is an unavoidable tension here. The government needs a rapid mechanism because regular cycles are slow, but a rapid mechanism must not reproduce the familiar pattern of approving technology before the operational problem has been fully understood.
TMF’s iterative and outcome-driven model offers some protection against that risk. Projects can be structured around phases and measurable results rather than receiving unrestricted funding based solely on a large initial promise.
Still, the compressed authorization window raises the stakes for selection. A September deadline can force decisions; it cannot make a weak project ready.
The most concrete implications are already clear:
The TMF Is Racing the Government’s Own Clock
As first reported by FedScoop, Jessie Posilkin, acting director of the Technology Modernization Fund, is being unusually explicit about both the targets and the urgency of the current proposal round. The fund wants projects that can speed and scale permitting work supported by the Council on Environmental Quality, as well as projects that can advance USAi adoption in agencies where capability or infrastructure gaps are blocking responsible use.Those priorities are broad enough to attract substantial interest but narrow enough to reveal the administration’s immediate technology agenda. Permitting modernization concerns the machinery through which government approves consequential physical projects; AI adoption concerns the machinery through which government increasingly expects its employees to analyze, write, code, and make decisions.
Both are bottleneck problems. Permitting can become trapped in fragmented systems, inconsistent data, manual handoffs, and weak coordination between agencies. AI can become trapped at the pilot stage because agencies lack prepared data, secure infrastructure, governance controls, acquisition pathways, or the capacity to operate tools beyond a demonstration.
The TMF’s official call says proposals outside the two priority areas remain welcome. In practice, however, agencies now know where the strongest policy signal lies, and they know they are competing against a calendar that leaves little space for vague concepts, unfinished business cases, or extended internal negotiation.
“TMF was made for moments like this, where agencies can’t afford to wait on budget and procurement cycles,” Posilkin wrote in announcing the push. That is an argument for the fund’s design, but it is also an indictment of the system around it: if normal budgeting cannot respond quickly enough to clearly identified technology needs, then the government’s emergency lane becomes indispensable.
The trouble is that the emergency lane is itself running out of road.
Two Priorities, One Underlying Modernization Problem
The two named tracks may look unrelated at first. One concerns the administrative process surrounding infrastructure and development; the other concerns the adoption of a fast-moving general-purpose technology.Yet the operational needs behind them overlap. Both require agencies to structure and share data, replace fragile manual workflows, build secure interfaces, coordinate across organizational boundaries, and define who is accountable when technology changes the way decisions are made.
| Priority track | Intended applicants | Problem TMF is targeting | Likely modernization emphasis |
|---|---|---|---|
| Speeding and scaling permitting | Agencies implementing Council on Environmental Quality-supported permitting projects | Slow, fragmented or difficult-to-scale permitting processes | Shared data, interagency coordination, expanded implementation and accelerated high-impact work |
| Advancing American AI | Agencies working with USAi but facing capability or infrastructure gaps | Barriers preventing responsible and broad AI adoption | Data readiness, secure infrastructure, pilots and guardrails, enterprise-scale AI capabilities |
The AI call is similarly oriented toward implementation rather than enthusiasm. It seeks high-impact, shovel-ready projects that can prepare data and infrastructure, pilot emerging tools while refining guardrails, or produce secure capabilities at enterprise scale.
That is a meaningful distinction. Federal agencies have no shortage of AI demonstrations, vendor briefings, experimentation groups, and employee interest; what many lack is the institutional substrate needed to turn a promising model into a reliable service.
The gap can include identity and access management, logging, data classification, records handling, privacy review, testing, procurement, workforce skills, integration with existing systems, and mechanisms for monitoring outputs after deployment. An AI pilot can be produced by a small team with limited data, but responsible agency-wide adoption requires controls that survive changes in personnel, vendors, models, and policy.
The TMF is therefore not simply offering to purchase AI tools. It is positioning itself to finance the less glamorous foundations without which those tools remain isolated experiments—or become unmanaged liabilities.
AI Makes Governance Harder, Not Optional
Posilkin’s warning about AI is more consequential than the funding announcement alone. “We are only seeing that the way that AI tooling is changing how people code is actually increasing some of the governance concerns and ability to manage information,” she said in remarks quoted by FedScoop.That observation cuts against the simplistic idea that AI automatically reduces the cost and complexity of government technology. Generative tools may help employees create code, documentation, queries, workflows, or analyses more quickly, but greater production speed can also multiply the number of artifacts an agency must review, secure, maintain, and eventually retire.
AI-assisted coding, for example, does not remove the need for software inventories, source control, dependency management, access restrictions, testing, security review, documentation, and ownership. It can instead produce more code across more teams, sometimes written by employees who could not have created the same software unaided and may be less prepared to recognize subtle defects.
The resulting challenge is not that AI necessarily generates bad work. It is that it changes the ratio between creation and oversight: agencies may suddenly be able to produce applications, scripts, analyses, and automations faster than their existing governance systems can classify and supervise them.
The same issue applies to information management. An employee who uses an AI tool to summarize documents, generate correspondence, analyze case files, or build an internal assistant may create new copies, derived records, prompts, logs, embeddings, indexes, and outputs. Each can carry retention, privacy, security, evidentiary, or public-disclosure implications.
Responsible adoption consequently requires more than a list of approved models. Agencies need enforceable rules about what information can enter a system, how outputs may be used, what activity must be logged, when human review is required, how results are challenged, and who has authority to shut down a deployment that behaves unexpectedly.
This is why the TMF’s emphasis on agencies with capability and infrastructure gaps matters. A weak foundation cannot be repaired merely by placing an AI interface over it. In some cases, AI will expose and amplify weaknesses that were already present in an agency’s data architecture, access controls, documentation, or software-development practices.
The best proposals will recognize that reality. They will describe AI as a governed service operating inside an accountable technical environment—not as a magic layer that compensates for everything underneath it.
Permitting Technology Has to Cross Agency Boundaries
The permitting track presents a different version of the same institutional challenge. A single permit or approval process may depend on several offices, systems, datasets, legal authorities, review stages, and external applicants, meaning that improvement cannot always be delivered by modernizing one application in isolation.A polished public portal can still conceal manual re-entry, emailed spreadsheets, inconsistent identifiers, and disconnected status systems behind the scenes. Applicants may receive a better-looking front end while government employees continue reconciling records by hand.
The TMF call’s emphasis on sharing data and coordinating with other agencies points toward the harder work. Effective modernization may require common data definitions, interoperable interfaces, reliable status exchange, ownership agreements, and processes that allow one agency to trust information supplied by another.
Technology can accelerate a review only when the underlying policy and operational handoffs are sufficiently clear to encode. If two agencies disagree over responsibility, required evidence, or the meaning of a status field, a new system may digitize that dispute rather than resolve it.
The strongest permitting proposals should therefore connect technical deliverables to measurable changes in the process. Faster document intake matters, but so do reduced duplicate submissions, fewer manual transfers, clearer status visibility, earlier identification of incomplete applications, and better coordination between reviewing bodies.
This is also where the TMF model can have value beyond a conventional appropriation. A modernization fund can support work that crosses organizational lines and does not fit cleanly into one office’s existing operations budget, particularly when the benefit is distributed among several agencies or accrues over multiple years.
But cross-agency projects are difficult to assemble quickly. They require sponsorship, agreements, architecture decisions, and clarity about who will operate the resulting service after the initial investment. The July 24 deadline leaves little time to create that alignment from scratch.
$200 Million Is Both Real Money and a Small Margin for Error
Posilkin said last month that the fund had $200 million remaining, pending potential reauthorization for fiscal 2027. In almost any individual agency budget discussion, that amount would represent substantial purchasing power.At government-wide scale, however, it is a constrained portfolio. It must be weighed against aging systems, cybersecurity requirements, citizen-service failures, data remediation, cloud migration, interoperability needs, and now an expanding set of AI infrastructure and governance demands.
“We are a country that continues to have really significant service delivery challenges, which is what the government should ultimately be providing,” Posilkin said. “And when you have it set up that way, $200 million is actually quite the drop in the bucket.”
Her point is not simply that government technology is expensive. It is that the cost of modernization is distributed unevenly and often becomes visible only when a legacy system fails, a security incident occurs, or a high-profile service cannot meet demand.
A fund of this size also cannot rescue every deserving project. Selection becomes an exercise in leverage: the board must identify investments that can unlock larger improvements, create reusable components, remove a particularly consequential bottleneck, or provide evidence that supports broader adoption.
That raises the cost of choosing poorly. A project that begins with an unclear problem, depends on unresolved policy decisions, or cannot survive beyond the initial funding period consumes capital that cannot be recovered merely by producing a demonstration.
The reference to 100% of agencies repaying TMF loans is therefore important but not sufficient. Repayment can sustain the revolving model, yet it does not eliminate the timing mismatch between repayments arriving and new needs emerging. Nor does it make the fund large enough to match the growing scale and scope of government technology challenges.
Repayment requirements can also influence which projects agencies bring forward. Projects with obvious savings or measurable cost avoidance may be easier to justify than public-service improvements whose benefits are substantial but difficult to convert into a repayment schedule.
This tension has followed modernization funding for years: government wants projects to demonstrate financial discipline, but some of its most important outcomes—less citizen frustration, more reliable access, shorter processing times, better security, and reduced operational risk—do not appear as simple revenue.
For the current call, agencies will need to connect mission value to a credible investment structure. A proposal cannot rely solely on the importance of AI or permitting as a policy area; it must explain why TMF financing is necessary, why the work is ready now, and how the investment produces durable value.
Congress Is Offering a Placeholder, Not Certainty
The current version of the fiscal 2027 Financial Services and General Government appropriations bill includes $5 million for the Technology Modernization Fund. The House committee report describes TMF as a revolving fund for digital transformation, larger multi-year technology upgrades, and urgent cybersecurity needs.But the bill has not received a floor vote in either chamber. Its treatment of the fund may change, and inclusion in committee legislation should not be mistaken for final reauthorization or a stable long-term mandate.
Even if enacted at the stated amount, $5 million would be minor next to the $200 million Posilkin said remained last month. It would signal that Congress had not entirely abandoned the mechanism, but it would not provide the scale implied by the administration’s ambitions for AI, permitting, cybersecurity, and citizen services.
The more immediate issue is authorization. The fund’s authority to make new investments ends on September 30 absent congressional action, leaving TMF staff to evaluate proposals, conduct due diligence, negotiate terms, and complete selections within a hard and rapidly approaching boundary.
That is why the July 24 submission date cannot be treated like an ordinary grant deadline. Waiting until the final day may technically preserve eligibility, but it reduces the opportunity to answer questions, resolve weaknesses, or adjust a proposal before the fund’s own decision window closes.
“This call is moving fast, partly out of necessity,” Posilkin wrote. “We have a very narrow window between now and September 30, when our authorization to make new investments ends (absent congressional action, of course…).”
Congress could extend that window. Until it does, agency leaders have to plan against the authority that exists, not the authority they hope will exist.
That distinction is familiar to federal IT executives. Long-term modernization is frequently expected to proceed through short-term legislative decisions, continuing uncertainty, and annual funding processes that do not align well with multi-year technical delivery.
The TMF was designed in part to escape that trap. Its current predicament shows that an alternative funding mechanism remains dependent on the same political calendar it was meant to help agencies navigate.
Timeline
July 2026 — TMF opens the focused call for proposals covering permitting technology and responsible AI adoption.July 24, 2026 — Agencies must submit initial project proposals for consideration under the call, with earlier submissions strongly encouraged.
September 30, 2026 — TMF’s authorization to make new investments ends unless Congress acts; AI proposals submitted under the call can be considered for selection and announcement by this date.
Fiscal 2027 — Potential reauthorization remains unresolved, while the current version of the relevant appropriations bill includes $5 million and awaits a floor vote in both chambers.
“Shovel-Ready” Is a Technical Standard, Not a Slogan
The accelerated schedule places unusual weight on proposal maturity. Agencies that have already documented a problem, identified an executive sponsor, mapped dependencies, estimated costs, and developed an acquisition strategy are positioned far better than teams still trying to define what they want to build.For an AI proposal, maturity should include a defined use case and the data necessary to support it. It should also include a realistic account of security, privacy, records, accessibility, evaluation, workforce impact, and human oversight.
A proposal that says an agency wants to “use AI to improve efficiency” is not a project plan. It does not identify the decision, workflow, service, dataset, user population, risk, or result that would allow reviewers to judge whether the investment is worthwhile.
A stronger proposal might describe a bounded operational process, explain why current systems or staffing cannot meet the need, and define what success would look like without presuming that model adoption is itself the desired outcome. It would separate the cost of the AI capability from the cost of preparing the agency to operate it responsibly.
Permitting proposals face an equivalent test. A project should identify the stage at which delay occurs, the agencies or parties involved, the information exchanged, and the technical or procedural constraint that prevents improvement.
Claims about acceleration require a baseline. If an agency cannot describe current processing time, manual effort, error rates, resubmissions, backlogs, or coordination delays, it may struggle to demonstrate that a new system produced a meaningful result.
The phrase shovel-ready also implies that projects cannot depend on a long chain of unresolved decisions after selection. TMF reviewers are likely to be wary of proposals requiring years of preliminary analysis before users or the public receive a benefit.
That does not mean agencies should hide complexity. The compressed window makes honest dependency mapping more valuable, because an optimistic schedule built on missing approvals or uncertain data access can fail faster than a cautious one.
The best candidate is not necessarily the project promising the most dramatic transformation. It is the project with a consequential problem, a credible delivery path, manageable dependencies, measurable outcomes, and leadership prepared to make decisions when implementation becomes difficult.
Agencies Must Prove They Can Operate What They Build
Technology modernization often receives attention at launch and loses it during operations. A new platform can be delivered successfully and still deteriorate if an agency has not funded maintenance, assigned ownership, retained expertise, monitored performance, or planned for changing requirements.AI makes that operational burden more visible. Models, integrations, policies, and usage patterns can change, requiring ongoing evaluation rather than a single acceptance test at deployment.
An enterprise AI capability may need monitoring for accuracy, drift, inappropriate access, misuse, cost escalation, and changes in the source data or model. If an agency lacks the people and processes to perform that work, the project is not truly ready for enterprise operation.
Permitting platforms have similarly persistent responsibilities. Data standards evolve, participating organizations change, regulations are updated, interfaces fail, user expectations rise, and the number of records grows.
TMF funding can help an agency overcome the initial investment barrier. It cannot substitute indefinitely for a product owner, operating budget, support model, security program, or governance structure.
Applicants should consequently treat sustainability as part of the architecture. They need to identify who owns the capability after the funded milestones, how ongoing costs will be covered, and whether the agency can maintain critical components without permanent dependence on one contractor.
The repayment condition reinforces this need. If an agency must both operate the service and repay the investment, an unrealistic cost model can undermine the project after the visible implementation phase has ended.
A credible application should distinguish among one-time modernization costs, recurring operational costs, repayment obligations, and expenses expected to be absorbed by existing programs. Blurring those categories may make a proposal appear cheaper, but it transfers risk into later budget years.
The Small TMF Team Is Part of the Constraint
Posilkin encouraged agencies to move quickly so the small TMF team can review proposals expeditiously. That detail deserves attention because application deadlines are only one half of a competitive funding process; review capacity is the other.A flood of late submissions can create pressure to triage quickly, prioritize proposals requiring less clarification, and spend limited staff time on projects that already present a coherent case. Agencies that submit early may give reviewers more opportunity to identify missing information while there is still time to respond.
This does not mean early submission guarantees selection. It means that proposal quality includes the ability to be understood and evaluated within the available window.
Federal technology applications can become dense compilations of policy language, broad benefits, architecture terminology, and optimistic milestones. Under a compressed review schedule, that style becomes a practical disadvantage.
The proposal needs to establish the problem, intervention, outcome, risk, cost, and implementation path without forcing reviewers to reconstruct the logic from attachments. It should be specific enough for technical scrutiny but clear enough for decision-makers to see why the project matters.
Applicants should also anticipate that the TMF team may challenge assumptions about scope, repayment, acquisition, security, and sequencing. A project team that cannot answer promptly because authority is scattered among several offices will lose time it does not have.
The July 24 date is therefore not the true beginning of the process. For serious applicants, work must begin well before submission, and internal approval channels should be prepared to handle follow-up at a pace that normal agency governance may not naturally support.
The Call Tests Whether Modernization Can Stay Bipartisan
Posilkin described the moment as both critical and bipartisan. That characterization is plausible because faster public services, more efficient permitting, secure technology, and better-managed AI do not belong exclusively to one political coalition.The harder question is whether agreement on outcomes will produce durable support for the financing mechanism. Congress can endorse modernization in principle while disagreeing over spending levels, executive priorities, oversight, repayment, or the size and role of a central fund.
The $5 million included in the current appropriations bill demonstrates the difference between recognition and scale. It acknowledges that TMF remains part of the government’s technology-financing framework, but it does not resolve how large or durable that role should be.
This matters because agencies make different choices when a program is perceived as temporary. They may rush proposals, narrow ambitions, defer cross-agency work, or avoid investing staff time in applications whose funding mechanism may disappear.
TMF staff face a similar problem. Building expertise in evaluating and overseeing complex modernization projects is itself a long-term institutional investment. Uncertain authority makes it harder to plan the portfolio, retain capability, and provide agencies with predictable guidance.
If lawmakers value the fund, the useful debate is not simply whether to keep its name in an appropriations bill. It is whether TMF should operate as an enduring government-wide investment mechanism with enough authority and capital to support a deliberate portfolio.
Without that certainty, every urgent call risks becoming a race to obligate funds before another deadline. That may produce successful projects, but it is an inefficient way to manage the government’s technology estate.
Action checklist for admins
- Identify permitting or AI projects that already have an executive sponsor, defined users, and a documented operational problem.
- Confirm that the project can submit an initial proposal no later than July 24, 2026, and aim to submit earlier.
- Separate essential enabling work—data, identity, infrastructure, security and governance—from optional product features.
- Document current performance so the project can demonstrate measurable improvement after investment.
- Map interagency, legal, acquisition and data-access dependencies before presenting the project as shovel-ready.
- Define long-term ownership, recurring operating costs, and a credible approach to TMF loan repayment.
- Prepare decision-makers and technical staff to answer reviewer questions quickly before the September 30 authorization deadline.
The July Call Will Reward Readiness Over Rhetoric
For agency CIOs and program leaders, the lesson is not to attach the word AI to an unfinished modernization request. The fund’s priorities may be politically prominent, but the accelerated schedule leaves little tolerance for projects that cannot connect policy urgency to executable work.Permitting teams should resist treating a new portal as the full solution when the actual delays lie in data exchange or organizational handoffs. AI teams should resist treating model access as enterprise adoption when the actual gaps concern governance, infrastructure, and operational ownership.
The TMF can supply capital and specialized oversight. It cannot manufacture consensus inside an agency, create a reliable dataset from undocumented records overnight, or assign accountability where leaders have avoided doing so.
That makes this round a useful test of federal technology maturity. It will show whether agencies have moved beyond broad modernization strategies and accumulated a pipeline of investment-ready projects that can be activated when financing becomes available.
It will also test the fund’s ability to preserve quality under deadline pressure. Moving quickly is valuable only if the resulting investments remain technically credible, governable, and sustainable.
There is an unavoidable tension here. The government needs a rapid mechanism because regular cycles are slow, but a rapid mechanism must not reproduce the familiar pattern of approving technology before the operational problem has been fully understood.
TMF’s iterative and outcome-driven model offers some protection against that risk. Projects can be structured around phases and measurable results rather than receiving unrestricted funding based solely on a large initial promise.
Still, the compressed authorization window raises the stakes for selection. A September deadline can force decisions; it cannot make a weak project ready.
What Agencies Should Carry Into the Deadline
The proposal call is ultimately about more than distributing the fund’s remaining resources. It is a practical demonstration of how the federal government intends to reconcile ambitious technology policy with limited capital, finite review capacity, and uncertain legislative authority.The most concrete implications are already clear:
- Initial project proposals are due by July 24, 2026, with earlier submissions strongly encouraged.
- TMF is prioritizing Council on Environmental Quality-supported permitting work and responsible USAi adoption.
- The fund had $200 million remaining as stated last month, but that money must address government-wide demand.
- TMF’s authority to make new investments ends on September 30, 2026, unless Congress acts.
- The current fiscal 2027 appropriations bill includes $5 million for TMF but has not received a floor vote in either chamber.
- Competitive projects will need mature plans for data, governance, delivery, operations, measurable outcomes, and repayment.