Microsoft announced on June 25, 2026, that Copilot in Excel is gaining reusable finance “Skills,” additional live financial data connectors, and more explicit change tracking for Microsoft 365 Copilot customers on Excel for the web, Windows, and Mac. The update is not just another AI button in the ribbon. It is Microsoft’s clearest attempt yet to make Excel less like a passive spreadsheet and more like a governed workspace for repeatable analytical labor. For finance teams, that distinction matters because the bottleneck has never been merely writing formulas; it has been trusting the process that produced them.
The most important part of this announcement is not that Copilot can now do more inside Excel. It is that Microsoft is trying to make Copilot do the same thing twice.
That is the promise behind Copilot Skills for Excel. A finance team can define a repeatable workflow — a discounted cash flow model, a variance analysis, a three-statement model, a close package, or a board-reporting template — and save the instructions as a
This is a meaningful shift in how Microsoft wants customers to think about AI in Office. Early Copilot demos were often built around the magic trick: ask a natural-language question, receive a generated chart, a formula, a summary, or a deck. Skills move the center of gravity away from improvisation and toward codified institutional knowledge.
That is exactly where Excel lives in real companies. The most valuable spreadsheet in an organization is rarely a clean greenfield workbook. It is usually a scarred, versioned, password-adjacent artifact filled with business rules, naming conventions, formatting habits, tribal assumptions, and one senior analyst who knows why row 147 must never be touched.
Skills are Microsoft’s attempt to bottle that context. The company is essentially saying that a finance department should not have to teach Copilot its preferred way of doing work every time a workbook opens.
That is why AI in Excel is both obvious and dangerous. Obvious, because spreadsheet work is full of repetitive translation: source data into model, model into summary, summary into presentation, presentation back into a revised forecast. Dangerous, because spreadsheets already hide too much logic behind cells, links, copied ranges, and aging assumptions.
Microsoft’s new finance push recognizes that Copilot cannot succeed in Excel by behaving like a chatbot floating beside the grid. It has to understand workbooks as governed business processes. It has to know not only what number to produce, but which sheet to touch, which formula pattern to preserve, which formatting convention matters, and which assumptions need to remain visible.
That is why the
Many organizations have spent years pretending their Excel processes are documented because the workbook exists. They are not. A workbook is an outcome, not a procedure. If Copilot Skills force teams to describe how their models should be built, reviewed, and reused, Microsoft may accidentally improve finance governance simply by making AI less tolerant of ambiguity.
A marketing team may tolerate a generated summary if it sounds plausible. A sales team may use AI to draft account notes and clean up language. A finance team, especially around close, audit, forecasting, treasury, or investor reporting, needs to know where a figure came from, what assumption changed, and whether the calculation can survive review.
That is why Microsoft is leaning hard into the language of traceability. The new Plan with Copilot mode is designed to show what Copilot intends to do before it does it. It can outline the cells, ranges, formulas, sheets, and assumptions it plans to modify, and it can ask clarifying questions before acting.
This matters because the most frightening AI failure in Excel is not a theatrical hallucination. It is a small, confident, hard-to-spot change in a model that looks normal until it makes its way into a forecast, a covenant calculation, a budget approval, or an executive slide.
Microsoft’s answer is not to claim that Copilot will be infallible. Instead, it is trying to make Copilot reviewable. That is the right instinct. In finance, trust is not created by a polished answer; it is created by a chain of custody.
This is the less glamorous but potentially more consequential part of the story. A large portion of financial analysis is not analysis at all. It is retrieval, reconciliation, and formatting: pulling market data, company fundamentals, private-market intelligence, filings data, ratings, estimates, or research from one system and getting it into the workbook without breaking the model.
Every finance professional knows the ritual. Open the terminal or portal. Export the data. Copy the table. Paste into Excel. Clean the fields. Check the dates. Fix the formatting. Wonder whether the number changed since the last pull. Repeat when the managing director asks for the same analysis with a different comp set.
Federated connectors are Microsoft’s attempt to compress that ritual. Instead of treating Excel as the endpoint of manually imported data, Copilot can retrieve information from approved providers at query time and bring it into the workbook. Some connectors will still require separate subscriptions, which is exactly what one would expect in institutional finance. Microsoft is not replacing the data vendors; it is trying to put them behind the same conversational and spreadsheet interface.
There is a practical upside here for Windows and Microsoft 365 shops. If the connector model works, Excel becomes a more defensible front end for licensed data. That means fewer ad hoc exports, fewer stale snapshots, and less incentive for analysts to paste sensitive or licensed data into unmanaged tools.
But there is also a new administrative burden. IT and finance operations teams will need to understand which connectors are enabled, which users can authenticate, what data can be retrieved, where it lands, and how that usage is logged. Live data inside AI-assisted workflows is useful only if entitlement, compliance, and audit controls keep pace.
Over the past two years, the market has filled with assistants promising to automate research, filings analysis, KPI extraction, variance commentary, private-company screening, and model generation. Some are startups. Some are add-ins. Some are large-language-model front ends wrapped around financial data sources. Some are impressive. Many are brittle.
Microsoft’s advantage is that it does not need to persuade finance teams to move into a new workspace. The work is already in Excel, Outlook, Teams, SharePoint, PowerPoint, and OneDrive. Copilot’s strategic job is to make that existing estate feel AI-native before a rival can make it feel obsolete.
That is why the finance connectors matter. If Copilot can pull from LSEG, Moody’s, Morningstar, PitchBook, FactSet, Daloopa, CB Insights, and S&P Global inside the Microsoft 365 workflow, the pitch to the CFO becomes much easier. Microsoft can say: keep your models, keep your permissions, keep your data vendors, keep your audit trail, and add AI where the work already happens.
This is classic platform strategy. Microsoft is not trying to own every data source or every workflow-specific finance product. It is trying to own the work surface where those sources and workflows converge.
For WindowsForum readers, the implication is straightforward. Excel is becoming one of the most important AI endpoints in the Microsoft stack. Not because spreadsheets are new, but because they are where AI’s claims about productivity collide with the hard reality of numbers that must be checked.
Microsoft says Copilot edits can be attributed separately from human collaborators in Excel’s change history. That distinction is critical. If a workbook says only that “something changed,” or if AI edits are blended into ordinary collaboration noise, the feature becomes a governance liability. If Copilot’s actions are clearly tagged, reviewable, and tied back to affected cells, the conversation changes.
This is where Microsoft is trying to thread a difficult needle. Users want Copilot to be agentic enough to modify workbooks, build structures, adjust formulas, and transform raw data. Organizations want the opposite guarantee: that nothing important changes invisibly.
Plan with Copilot and Copilot attribution are Microsoft’s compromise. Copilot can act, but it should preview its intended action and leave tracks afterward. That is not a complete control framework, but it is the minimum viable trust layer for AI in a spreadsheet.
Admins should still be cautious. An audit trail after the fact does not eliminate the need for workbook protection, data-loss prevention, sensitivity labels, versioning, approval workflows, and human review. Copilot does not turn a fragile spreadsheet into a controlled financial system. It makes some spreadsheet work more observable, which is useful but not magic.
Macros automated actions. Skills describe intent, standards, and workflow. A macro might run a fixed sequence of operations. A Skill tells Copilot how to approach a class of work, what structure to use, and what output should look like. That makes Skills more flexible, but also harder to reason about.
The old macro world had its own problems: security warnings, unsigned code, brittle VBA, personal workbooks full of undocumented automation, and business-critical processes maintained by whoever had not yet left the company. Skills could recreate some of that risk in a new language if organizations treat them casually.
A
This is where IT should get ahead of the curve. Organizations will need conventions for naming, storing, approving, versioning, and retiring Skills. Finance leaders will need to decide which Skills are personal productivity aids and which are official departmental procedures. Security teams will need to understand whether Skills can be used to steer Copilot into retrieving or transforming sensitive data in ways that violate policy.
The lesson from macros is not that automation should be avoided. It is that business users will automate whether IT blesses it or not. The smarter move is to give them a governed path before the shadow version becomes the default.
That rollout pattern is worth parsing. The core controls and prebuilt capabilities are not being positioned as distant research previews. Microsoft is putting them in front of paying Copilot customers now, while holding back custom Skills for a more gradual release.
That suggests Microsoft understands the risk profile. Prebuilt Skills and governed connectors are easier to support. Custom Skills are where organizations will discover both the power and the weirdness of reusable AI instructions. The Insiders-first path gives Microsoft a buffer while users test real workflows and inevitably find edge cases.
It also means admins should not wait until every feature is broadly available to start planning. If Microsoft 365 Copilot is already deployed, finance users may begin experimenting quickly. The question for IT is not whether someone will try to make Copilot close the month. The question is whether the organization will have any policy in place before they do.
Finance teams already live with wrong spreadsheets. They live with broken links, hidden rows, stale assumptions, circular references, copied formulas that skip a row, and workbooks named “final_v7_revised_REALfinal.xlsx.” The question is not whether AI introduces error into an error-free environment. It is whether AI makes errors easier to detect or easier to disguise.
Microsoft’s new features cut both ways. Skills can standardize good practice, but they can also standardize a bad assumption. Connectors can reduce stale copy-paste work, but they can also make users less aware of the provenance and limitations of the data being retrieved. Plan with Copilot can improve review, but only if users actually read the plan rather than clicking through because the quarter-end clock is ticking.
This is why the cultural adoption problem matters as much as the product design. A finance organization that treats Copilot as a junior analyst with a visible workpaper trail may get real value. A finance organization that treats Copilot as an oracle will eventually get burned.
The right mental model is not “AI replaces Excel work.” It is “AI changes where review should happen.” Instead of checking every mechanical step after hours of manual work, teams may review the Skill, the data source, the proposed plan, the changed cells, and the final output. That can be faster and more controlled, but only if the review discipline moves with the workflow.
Excel is not just another app in the suite. It is often a semi-official system of record, a reporting layer, a modeling environment, and an integration workaround. When Copilot gains the ability to act inside Excel with reusable Skills and live data connectors, admins need to understand the operational blast radius.
The immediate checklist is familiar but more urgent: licensing, identity, conditional access, data boundaries, sensitivity labels, audit logs, connector permissions, and user education. The harder work is organizational. IT will need a relationship with finance power users that goes beyond telling them not to paste confidential data into random chatbots.
The best governance model will probably be federated. Central IT should control platform policy, connector approval, security posture, and monitoring. Finance operations should own the content of approved Skills and the business logic behind them. Internal audit and compliance should help define which workflows require review, documentation, or retention.
That sounds bureaucratic, but it is the price of turning AI from a novelty into infrastructure. Once Copilot starts participating in financial models, the “who approved this?” question becomes unavoidable.
The announcement is full of features aimed at messiness: reusable instructions for recurring workflows, data connectors for trusted sources, planning before edits, and change attribution afterward. That is a much more credible direction than simply promising that Copilot can generate formulas or summarize tables.
Excel’s messiness is not a defect Microsoft can remove without destroying the thing people use it for. The grid is popular because it is flexible enough to accommodate incomplete data, provisional assumptions, one-off analyses, and business judgment. The challenge is to add AI without turning that flexibility into an accountability vacuum.
Reusable Skills are the strongest expression of that strategy. They acknowledge that the value of finance work is not just the output, but the method. A discounted cash flow model is not useful merely because it has a terminal value cell. It is useful because everyone involved understands the assumptions, layout, source data, and review process behind it.
If Copilot can help preserve that method while reducing the repetitive labor around it, Microsoft has something substantial. If it merely generates prettier workbooks faster, the novelty will fade.
Microsoft’s latest Copilot in Excel update is not the end of spreadsheet work; it is the next phase of spreadsheet institutionalization. The grid remains, the formulas remain, and the human judgment remains, but the repeatable mechanics around finance work are starting to move into reusable AI procedures and live data pipes. If Microsoft can keep the controls visible and the ecosystem governed, Excel may become the place where enterprise AI stops performing demos and starts doing accountable work.
Microsoft Is Turning Prompting Into Procedure
The most important part of this announcement is not that Copilot can now do more inside Excel. It is that Microsoft is trying to make Copilot do the same thing twice.That is the promise behind Copilot Skills for Excel. A finance team can define a repeatable workflow — a discounted cash flow model, a variance analysis, a three-statement model, a close package, or a board-reporting template — and save the instructions as a
SKILL.md file in OneDrive. Copilot can then use that file as a standing operating procedure rather than waiting for an analyst to reconstruct the same giant prompt every month.This is a meaningful shift in how Microsoft wants customers to think about AI in Office. Early Copilot demos were often built around the magic trick: ask a natural-language question, receive a generated chart, a formula, a summary, or a deck. Skills move the center of gravity away from improvisation and toward codified institutional knowledge.
That is exactly where Excel lives in real companies. The most valuable spreadsheet in an organization is rarely a clean greenfield workbook. It is usually a scarred, versioned, password-adjacent artifact filled with business rules, naming conventions, formatting habits, tribal assumptions, and one senior analyst who knows why row 147 must never be touched.
Skills are Microsoft’s attempt to bottle that context. The company is essentially saying that a finance department should not have to teach Copilot its preferred way of doing work every time a workbook opens.
The Spreadsheet Was Always a Workflow Engine in Disguise
Excel’s genius has never been that it is the best database, the best BI tool, the best programming environment, or the best document editor. It is that it is good enough at all of them to become the place where business reality gets negotiated.That is why AI in Excel is both obvious and dangerous. Obvious, because spreadsheet work is full of repetitive translation: source data into model, model into summary, summary into presentation, presentation back into a revised forecast. Dangerous, because spreadsheets already hide too much logic behind cells, links, copied ranges, and aging assumptions.
Microsoft’s new finance push recognizes that Copilot cannot succeed in Excel by behaving like a chatbot floating beside the grid. It has to understand workbooks as governed business processes. It has to know not only what number to produce, but which sheet to touch, which formula pattern to preserve, which formatting convention matters, and which assumptions need to remain visible.
That is why the
SKILL.md approach is interesting. Markdown is simple, portable, and readable by humans. It does not require finance teams to become software engineers, but it nudges them toward writing their process down in a structured way. In practice, that may be as important as the AI feature itself.Many organizations have spent years pretending their Excel processes are documented because the workbook exists. They are not. A workbook is an outcome, not a procedure. If Copilot Skills force teams to describe how their models should be built, reviewed, and reused, Microsoft may accidentally improve finance governance simply by making AI less tolerant of ambiguity.
Finance Is the Right Battlefield Because Finance Has Less Patience for Vibes
Microsoft’s choice of finance as the showcase is not accidental. Finance teams are spreadsheet maximalists, but they are also professionally allergic to unexplained numbers.A marketing team may tolerate a generated summary if it sounds plausible. A sales team may use AI to draft account notes and clean up language. A finance team, especially around close, audit, forecasting, treasury, or investor reporting, needs to know where a figure came from, what assumption changed, and whether the calculation can survive review.
That is why Microsoft is leaning hard into the language of traceability. The new Plan with Copilot mode is designed to show what Copilot intends to do before it does it. It can outline the cells, ranges, formulas, sheets, and assumptions it plans to modify, and it can ask clarifying questions before acting.
This matters because the most frightening AI failure in Excel is not a theatrical hallucination. It is a small, confident, hard-to-spot change in a model that looks normal until it makes its way into a forecast, a covenant calculation, a budget approval, or an executive slide.
Microsoft’s answer is not to claim that Copilot will be infallible. Instead, it is trying to make Copilot reviewable. That is the right instinct. In finance, trust is not created by a polished answer; it is created by a chain of custody.
Live Data Connectors Attack the Copy-Paste Economy
The other major piece of the update is data access. Microsoft says Copilot in Excel is adding financial data connectors from CB Insights, Daloopa, FactSet, Morningstar, PitchBook, and S&P Global. These join LSEG and Moody’s connectors that arrived earlier in the year.This is the less glamorous but potentially more consequential part of the story. A large portion of financial analysis is not analysis at all. It is retrieval, reconciliation, and formatting: pulling market data, company fundamentals, private-market intelligence, filings data, ratings, estimates, or research from one system and getting it into the workbook without breaking the model.
Every finance professional knows the ritual. Open the terminal or portal. Export the data. Copy the table. Paste into Excel. Clean the fields. Check the dates. Fix the formatting. Wonder whether the number changed since the last pull. Repeat when the managing director asks for the same analysis with a different comp set.
Federated connectors are Microsoft’s attempt to compress that ritual. Instead of treating Excel as the endpoint of manually imported data, Copilot can retrieve information from approved providers at query time and bring it into the workbook. Some connectors will still require separate subscriptions, which is exactly what one would expect in institutional finance. Microsoft is not replacing the data vendors; it is trying to put them behind the same conversational and spreadsheet interface.
There is a practical upside here for Windows and Microsoft 365 shops. If the connector model works, Excel becomes a more defensible front end for licensed data. That means fewer ad hoc exports, fewer stale snapshots, and less incentive for analysts to paste sensitive or licensed data into unmanaged tools.
But there is also a new administrative burden. IT and finance operations teams will need to understand which connectors are enabled, which users can authenticate, what data can be retrieved, where it lands, and how that usage is logged. Live data inside AI-assisted workflows is useful only if entitlement, compliance, and audit controls keep pace.
Microsoft Is Building a Moat Around the Finance Desktop
There is a competitive story hiding underneath this announcement. Microsoft is not merely adding features to Excel; it is defending Excel against a new class of AI-native finance tools.Over the past two years, the market has filled with assistants promising to automate research, filings analysis, KPI extraction, variance commentary, private-company screening, and model generation. Some are startups. Some are add-ins. Some are large-language-model front ends wrapped around financial data sources. Some are impressive. Many are brittle.
Microsoft’s advantage is that it does not need to persuade finance teams to move into a new workspace. The work is already in Excel, Outlook, Teams, SharePoint, PowerPoint, and OneDrive. Copilot’s strategic job is to make that existing estate feel AI-native before a rival can make it feel obsolete.
That is why the finance connectors matter. If Copilot can pull from LSEG, Moody’s, Morningstar, PitchBook, FactSet, Daloopa, CB Insights, and S&P Global inside the Microsoft 365 workflow, the pitch to the CFO becomes much easier. Microsoft can say: keep your models, keep your permissions, keep your data vendors, keep your audit trail, and add AI where the work already happens.
This is classic platform strategy. Microsoft is not trying to own every data source or every workflow-specific finance product. It is trying to own the work surface where those sources and workflows converge.
For WindowsForum readers, the implication is straightforward. Excel is becoming one of the most important AI endpoints in the Microsoft stack. Not because spreadsheets are new, but because they are where AI’s claims about productivity collide with the hard reality of numbers that must be checked.
The Audit Trail Is the Product, Not the Fine Print
The Show Changes pane is not the flashiest part of the announcement, but it may be the part that decides whether finance teams actually use these features.Microsoft says Copilot edits can be attributed separately from human collaborators in Excel’s change history. That distinction is critical. If a workbook says only that “something changed,” or if AI edits are blended into ordinary collaboration noise, the feature becomes a governance liability. If Copilot’s actions are clearly tagged, reviewable, and tied back to affected cells, the conversation changes.
This is where Microsoft is trying to thread a difficult needle. Users want Copilot to be agentic enough to modify workbooks, build structures, adjust formulas, and transform raw data. Organizations want the opposite guarantee: that nothing important changes invisibly.
Plan with Copilot and Copilot attribution are Microsoft’s compromise. Copilot can act, but it should preview its intended action and leave tracks afterward. That is not a complete control framework, but it is the minimum viable trust layer for AI in a spreadsheet.
Admins should still be cautious. An audit trail after the fact does not eliminate the need for workbook protection, data-loss prevention, sensitivity labels, versioning, approval workflows, and human review. Copilot does not turn a fragile spreadsheet into a controlled financial system. It makes some spreadsheet work more observable, which is useful but not magic.
The Skill File May Become the New Macro
There is a tempting comparison between Copilot Skills and Excel macros. It is not perfect, but it is revealing.Macros automated actions. Skills describe intent, standards, and workflow. A macro might run a fixed sequence of operations. A Skill tells Copilot how to approach a class of work, what structure to use, and what output should look like. That makes Skills more flexible, but also harder to reason about.
The old macro world had its own problems: security warnings, unsigned code, brittle VBA, personal workbooks full of undocumented automation, and business-critical processes maintained by whoever had not yet left the company. Skills could recreate some of that risk in a new language if organizations treat them casually.
A
SKILL.md file in OneDrive sounds friendly, and that is part of the appeal. But reusable AI instructions are still operational assets. They can encode assumptions, process steps, formatting rules, and analytical methods. If they are wrong, outdated, or too broadly shared, the mistake can scale.This is where IT should get ahead of the curve. Organizations will need conventions for naming, storing, approving, versioning, and retiring Skills. Finance leaders will need to decide which Skills are personal productivity aids and which are official departmental procedures. Security teams will need to understand whether Skills can be used to steer Copilot into retrieving or transforming sensitive data in ways that violate policy.
The lesson from macros is not that automation should be avoided. It is that business users will automate whether IT blesses it or not. The smarter move is to give them a governed path before the shadow version becomes the default.
Availability Tells Us This Is More Than a Demo
Microsoft says personalization, workbook rules, prebuilt finance Skills, federated Copilot connectors, Plan with Copilot, and Copilot attribution in Show Changes are generally available for Microsoft 365 Copilot customers using Excel for the web, Windows, and Mac. Custom Skills are available through the Insiders channel for Windows and Mac and are expected to reach general availability over the next month. Partner-built Skills from companies including LSEG, Ramp, Rogo, and Vena are expected later in 2026 through Microsoft Marketplace.That rollout pattern is worth parsing. The core controls and prebuilt capabilities are not being positioned as distant research previews. Microsoft is putting them in front of paying Copilot customers now, while holding back custom Skills for a more gradual release.
That suggests Microsoft understands the risk profile. Prebuilt Skills and governed connectors are easier to support. Custom Skills are where organizations will discover both the power and the weirdness of reusable AI instructions. The Insiders-first path gives Microsoft a buffer while users test real workflows and inevitably find edge cases.
It also means admins should not wait until every feature is broadly available to start planning. If Microsoft 365 Copilot is already deployed, finance users may begin experimenting quickly. The question for IT is not whether someone will try to make Copilot close the month. The question is whether the organization will have any policy in place before they do.
The Real Risk Is Overconfidence, Not Automation
The common critique of AI in spreadsheets is that the model might be wrong. That is true, but incomplete.Finance teams already live with wrong spreadsheets. They live with broken links, hidden rows, stale assumptions, circular references, copied formulas that skip a row, and workbooks named “final_v7_revised_REALfinal.xlsx.” The question is not whether AI introduces error into an error-free environment. It is whether AI makes errors easier to detect or easier to disguise.
Microsoft’s new features cut both ways. Skills can standardize good practice, but they can also standardize a bad assumption. Connectors can reduce stale copy-paste work, but they can also make users less aware of the provenance and limitations of the data being retrieved. Plan with Copilot can improve review, but only if users actually read the plan rather than clicking through because the quarter-end clock is ticking.
This is why the cultural adoption problem matters as much as the product design. A finance organization that treats Copilot as a junior analyst with a visible workpaper trail may get real value. A finance organization that treats Copilot as an oracle will eventually get burned.
The right mental model is not “AI replaces Excel work.” It is “AI changes where review should happen.” Instead of checking every mechanical step after hours of manual work, teams may review the Skill, the data source, the proposed plan, the changed cells, and the final output. That can be faster and more controlled, but only if the review discipline moves with the workflow.
Windows Admins Inherit Another Business-Critical AI Surface
For sysadmins and Microsoft 365 administrators, this announcement is a reminder that Copilot governance is no longer an abstract tenant-level topic. It is moving into specific applications, specific roles, and specific business processes.Excel is not just another app in the suite. It is often a semi-official system of record, a reporting layer, a modeling environment, and an integration workaround. When Copilot gains the ability to act inside Excel with reusable Skills and live data connectors, admins need to understand the operational blast radius.
The immediate checklist is familiar but more urgent: licensing, identity, conditional access, data boundaries, sensitivity labels, audit logs, connector permissions, and user education. The harder work is organizational. IT will need a relationship with finance power users that goes beyond telling them not to paste confidential data into random chatbots.
The best governance model will probably be federated. Central IT should control platform policy, connector approval, security posture, and monitoring. Finance operations should own the content of approved Skills and the business logic behind them. Internal audit and compliance should help define which workflows require review, documentation, or retention.
That sounds bureaucratic, but it is the price of turning AI from a novelty into infrastructure. Once Copilot starts participating in financial models, the “who approved this?” question becomes unavoidable.
Microsoft’s Finance Copilot Is Useful Because It Admits Excel Is Messy
The most encouraging thing about this update is that it does not pretend finance work is clean.The announcement is full of features aimed at messiness: reusable instructions for recurring workflows, data connectors for trusted sources, planning before edits, and change attribution afterward. That is a much more credible direction than simply promising that Copilot can generate formulas or summarize tables.
Excel’s messiness is not a defect Microsoft can remove without destroying the thing people use it for. The grid is popular because it is flexible enough to accommodate incomplete data, provisional assumptions, one-off analyses, and business judgment. The challenge is to add AI without turning that flexibility into an accountability vacuum.
Reusable Skills are the strongest expression of that strategy. They acknowledge that the value of finance work is not just the output, but the method. A discounted cash flow model is not useful merely because it has a terminal value cell. It is useful because everyone involved understands the assumptions, layout, source data, and review process behind it.
If Copilot can help preserve that method while reducing the repetitive labor around it, Microsoft has something substantial. If it merely generates prettier workbooks faster, the novelty will fade.
The Numbers Now Need a Chain of Custody
The practical lesson for finance teams is that Copilot in Excel should be introduced like a process change, not like a new autocomplete feature. The organizations that get the most from this update will be the ones that treat Skills, connectors, and change tracking as parts of the same control story.- Teams should begin with low-risk recurring workflows where the process is well understood and the output is easy to validate.
- Custom Skills should be versioned, reviewed, and owned by a named business function rather than left as personal prompt files.
- Financial data connectors should be approved with the same seriousness as any other source of licensed, sensitive, or decision-critical data.
- Copilot’s planned edits should be reviewed before execution, especially in workbooks used for reporting, forecasting, compliance, or executive decisions.
- Change attribution should become part of the review habit, not a forensic tool used only after something goes wrong.
- Administrators should assume that finance users will adopt these tools quickly if they reduce manual work, and governance should arrive before informal practices harden.
Microsoft’s latest Copilot in Excel update is not the end of spreadsheet work; it is the next phase of spreadsheet institutionalization. The grid remains, the formulas remain, and the human judgment remains, but the repeatable mechanics around finance work are starting to move into reusable AI procedures and live data pipes. If Microsoft can keep the controls visible and the ecosystem governed, Excel may become the place where enterprise AI stops performing demos and starts doing accountable work.
References
- Primary source: Tech My Money
Published: Sat, 27 Jun 2026 00:23:53 GMT
Loading…
techmymoney.com - Related coverage: daloopa.com
Daloopa expands financial data access with new MCP connector for Microsoft 365 Copilot integration, bringing trusted financial data directly into AI workflows - Daloopa
Enables investment firms to use LLMs with high-quality, AI-ready data for accurate analysisdaloopa.com - Related coverage: finance.sina.com.cn
Loading…
finance.sina.com.cn - Related coverage: natlawreview.com
CB Insights Brings Private Market Intelligence to Microsoft 365 Copilot and Copilot in Excel
Official Microsoft connector brings CB Insights’ trusted intelligence on private companies directly into AI-powered enterprise workflowsnatlawreview.com
- Official source: techcommunity.microsoft.com
Introducing federated Copilot connectors for LSEG and Moody's in Excel | Microsoft Community Hub
Earlier this month, we announced Microsoft 365 Copilot federated connectors were coming to Copilot in Excel. Built on the emerging industry standard...
techcommunity.microsoft.com
- Related coverage: fintech-pulse.com
Loading…
fintech-pulse.com
- Related coverage: au.investing.com
Microsoft’s Copilot in Excel adds skills and data connectors for finance work By Investing.com
Microsoft’s Copilot in Excel adds skills and data connectors for finance workau.investing.com - Related coverage: m.in.investing.com
Microsoft’s Copilot in Excel adds skills and data connectors for finance work By Investing.com
Microsoft’s Copilot in Excel adds skills and data connectors for finance workm.in.investing.com - Official source: news.microsoft.com
Copilot en Excel: Creado para la era de Frontier Finance - Source LATAM
news.microsoft.com
- Related coverage: sahmcapital.com
Loading…
www.sahmcapital.com - Related coverage: thurrott.com
Microsoft Excel Adds New Financial Connectors and Reusable Copilot Skills
Microsoft announced today the introduction of new financial connectors in Microsoft Excel, along with reusable Copilot Skills.www.thurrott.com - Related coverage: resultsense.com
Loading…
www.resultsense.com - Related coverage: press.spglobal.com
S&P Global Launches AI-Ready Commodities Data Integrated With Microsoft 365 Copilot - Apr 21, 2025
PDF documentpress.spglobal.com
- Official source: cdn-dynmedia-1.microsoft.com
Loading…
cdn-dynmedia-1.microsoft.com - Related coverage: benchmarkitservices.com
Loading…
www.benchmarkitservices.com - Official source: microsoft.com
Frontier Program Features | Microsoft Frontier
Explore the official inventory of Microsoft Frontier features. Track experimental agent capabilities as they evolve from early access to general release.www.microsoft.com
- Official source: learn.microsoft.com
Overview of Finance agents in Microsoft 365 2026 release wave 1 | Microsoft Learn
Overview of Finance agents in Microsoft 365 2026 release wave 1learn.microsoft.com - Official source: fpc.microsoft.com
Loading…
fpc.microsoft.com - Official source: adoption.microsoft.com