Roman Kepczyk’s annual technology forecast for accounting firms lays out a tight, practitioner-focused set of predictions that should be on every firm leader’s 2026 strategic radar: cloud engagement binder stagnation, change management as a competitive moat, vendor‑led adoption of agentic AI, rising pressure on PDF/OCR licensing, a contraction in AI startups, the rapid ascent of Copilot‑powered analytics in finance, renewed scrutiny of legacy document management, collaboration overtaking email, and a hard pivot toward managed, accounting‑specific cybersecurity and fully digital client ingress.
Background / Overview
The accounting profession entered 2024–2025 in a period of pragmatic experimentation: generative AI moved from novelty to a required component of vendor roadmaps, and a new class of agentic or multi‑agent systems began surfacing in products aimed squarely at bookkeeping, tax, and audit workflows. Kepczyk’s piece synthesizes this vendor momentum with firm‑level realities — training gaps, migration friction, and risk concerns — and makes ten practical predictions for where firms should prioritize resources in 2026. His framing reflects two consistent themes visible across industry reporting: vendors are productizing agentic features and Microsoft’s Fabric + Copilot story is reshaping how finance teams consume analytics. This feature unpacks each prediction, verifies central technical claims where public data exists, flags unverifiable or speculative points, and provides tactical guidance firms can use to convert predictions into measurable strategic actions.Engagement binders: migration stalls, but pockets of change continue
Kepczyk predicts continued stasis on wholesale migration from on‑premise engagement binders to cloud‑native assurance and tax binder solutions, citing feature parity and poor conversion tooling as friction points. This is consistent with vendor messaging and practitioner anecdotes: firms migrate when the ROI and migration risk align, and many midsize practices remain risk‑averse on record‑keeping infrastructure.Why the stall matters
- Engagement binders are the core record of attest work; migration affects audit trail integrity, retention policies, and automation opportunities.
- Vendors continue to invest in cloud engagement features, but migration tooling (bulk reindexing, metadata mapping, evidence provenance transfer) remains the gating factor.
- Look for incremental wins: hybrid approaches that retain legacy evidence while provisioning cloud‑native metadata overlays and indexing.
- Test conversions on a single engagement class and demand event‑level audit trails from vendors before any broad migration.
Change management becomes a superpower — invest in people now
One of the clearest, repeatedly validated predictions is that change management will determine winners and losers. Firms adopting a best‑of‑breed stack must accelerate adoption cycles, retrain staff, and redesign workflows. The failure modes for AI and automation projects are rarely purely technical; they’re organizational.Actions that pay off
- Make structured training a line‑item in the 2026 budget (hands‑on sandboxing, prompt hygiene, verification checklists).
- Adopt “learning organization” practices: short feedback loops, build‑measure‑learn pilots, and a rollback playbook.
- Rework job descriptions and career ladders to include AgentOps / automation verification responsibilities.
Vendors lead agentic AI adoption — expect Intuit and other incumbents to push “agents”
Kepczyk expects the first agentic systems adopted by firms to arrive packaged by major accounting vendors rather than through DIY or niche startups. Intuit’s public positioning on “agentic AI” confirms this trend: Intuit announced agentic capabilities and a GenOS framework designed to deliver done‑for‑you workflows across TurboTax, QuickBooks and other properties, with staged rollouts beginning in 2025. That vendor‑led path lowers integration risk and wraps agent behavior inside contractual and compliance frameworks firms prefer. How firms should respond- Treat vendor agents as production software: require SOC/ISO reports, test datasets, rollback controls, and event‑level logs.
- Insist on human‑in‑the‑loop approvals for any material outcomes (journal entries, tax positions).
- Negotiate clear data residency and “no‑train” clauses where client confidentiality is at stake.
AI data ingestion threatens PDF/OCR incumbents — but the transition is uneven
Kepczyk cites a statistic — that 66% of firms use OCR to capture tax source documents and that many pay for Adobe Acrobat/Adobe DC — to argue AI‑based extraction will put PDF vendors on notice. While Adobe Acrobat remains widely used for PDF workflows, the exact 66% figure referenced in the original commentary could not be independently verified through public surveys accessible during reporting; treat that specific number with caution.What is verifiable and meaningful
- Modern AI extraction and document‑to‑ledger tools have improved rapidly; specialized vendors (ledger overlays, invoice capture engines, document extraction startups) now produce structured outputs that rival or exceed legacy OCR pipelines on certain tax and bookkeeping tasks.
- Firms frequently pay sizeable PDF licensing fees; a shift to AI‑first extraction could reduce per‑seat costs — but only after careful validation against firm sample documents.
- Select a representative sample of recurring tax packages and run parallel ingestion through your Adobe/DC workflow and an AI extraction pilot.
- Measure extraction accuracy, exception rates, throughput, and total cost of ownership (including human review time).
- Map false positives/negatives into risk tolerance levels for tax return assembly.
AI startup bubble and vendor consolidation — risk and opportunity
Kepczyk forecasts a significant vendor shakeout in 2026 as private equity and venture funding pressures force underperforming AI startups to fold or be acquired. Independent industry surveys and analyst commentary echo the same reality: as 2024–2025 matured, abandonment rates for AI initiatives and skepticism around ROI rose, creating a clearing where only solutions with clear product‑market fit, auditability, and predictable economics survive. RAND, S&P Global Market Intelligence and other analyst summaries have reported high failure/abandonment rates for AI projects and warned of a market correction. Implications for buyers- Favor vendors that can demonstrate reproducible accuracy on representative firm datasets and provide contractual protections for model‑drift and SLA failures.
- Maintain a watchlist of startup IP and M&A activity — consolidation can be an opportunity to acquire advanced capabilities through established partners, but beware integration risk.
Copilot in Fabric and Power BI: finance teams will be surprised — Power BI is poised to dominate reporting
Kepczyk suggests that Copilot in Microsoft Fabric will raise finance teams’ expectations and push Power BI to dominance as the reporting tool of choice for accounting firms. Microsoft’s Fabric + Copilot rollout — including Copilot experiences across Fabric workloads and a direct Power BI Copilot integration — is publicly documented and has been broadly promoted as a step‑change for conversational analytics and automated query translation (natural language → KQL/DAX). Microsoft’s documentation and product blog confirm Copilot in Fabric for Power BI and the availability of AI skills and Q&A experiences that translate conversational questions into queries. What firms should do now- Pilot Copilot in Fabric on internal financials first: measure time‑to‑insight, accuracy of generated queries, and governance controls.
- Build a semantic model (a governed Power BI dataset) as a single source of truth before layering Copilot.
- Train finance teams to validate Copilot outputs — treat auto‑generated narratives as first drafts.
Legacy document management will be challenged — expect Microsoft + vertical players to respond
Kepczyk forecasts at least one major accounting vendor will roll out a DMS with embedded AI query capabilities — likely leveraging Microsoft SharePoint given the profession’s Microsoft‑centric stack. This aligns with vendor movements: Intapp and others are tightly integrating with Microsoft 365 to deliver engagement‑centric workspaces and document governance on top of SharePoint and Teams. Those integrations position SharePoint as the substrate for modern DMS plus AI indexing and search. Practical considerations- Audit your document metadata strategy now — AI search and Q&A are only effective when content is discoverable and tagged consistently.
- Consider an incremental transition: map retention/retention holds, redaction workflows, and eDiscovery requirements before decommissioning legacy DMS.
Email fades as primary collaboration — Teams, practice management systems and portals will take center stage
Kepczyk’s analogy comparing email’s decline to the old “pink slip” phone messages is a practical way to describe a real operational shift: enterprises and firms are moving toward threaded, document‑linked collaboration in Teams, shared workspaces, and practice management tools. Microsoft Teams’ massive user base and integration with Microsoft 365 makes it an obvious candidate for replacing ad‑hoc email collaboration in firms; published counts and market analyses document hundreds of millions of Teams users and rapid adoption among enterprise customers. How to migrate communication hygiene- Define the scope of Teams vs email for client communications (external vs internal).
- Build templates and governance—channel naming, retention, information barriers—and archive external correspondence for compliance.
- Evaluate practice management systems that can ingest client email and convert it into structured tasks and records.
Dark AI and cybersecurity: elevate to managed, verticalized defenses
Kepczyk warns that cybercriminals are using AI to scale attacks and that traditional cybersecurity governance will fail, driving most firms to accounting‑specific managed security services. The underlying trend is supported by security vendors and analyst commentary: generative tools have lowered the bar for sophisticated phishing, automated malware generation, and social engineering. Given firms’ constrained security staff and regulated data, specialized managed cybersecurity providers with accounting‑aware playbooks offer a pragmatic option.Recommended steps
- Treat vendor/agent connectors as high‑risk attack surfaces; require ephemeral credentials, least privilege, and connector‑level MFA.
- Move to an MSSP or a managed SOC that understands CPA‑firm controls, billing‑season risk spikes, and regulatory logging needs.
- Add model‑risk and AI‑use clauses to cyber insurance and vendor contracts.
Client ingress automation: standardization wins peak season
Finally, Kepczyk predicts that 2026 will be the year where the majority of client documents arrive via standardized electronic ingress tools — client portals, structured e‑forms, API feeds — rather than ad‑hoc file transfers. The logic is straightforward: standardized ingress minimizes rekeying, reduces OCR exceptions, and creates the structured inputs necessary for agentic workflows and automated reconciliation.Tactical rollout
- Prioritize a single client intake format (e.g., a standardized zipped package or API‑backed portal) for top 50 clients as a pilot.
- Measure exceptions, time‑to‑process, and client satisfaction; refine and scale iteratively.
- Use the gains to reprice seasonal work and free capacity for advisory services.
Retrospective: how last year’s predictions fared — lessons for 2026 planning
Kepczyk’s self‑assessment of 2025 predictions shows a mixed hit‑rate: wins on generative AI adoption and cloud migration, draws on tax research startups, and misses where adoption required deeper organizational change (skill transformation). That pattern is instructive: vendor features and product availability often arrive faster than firms’ ability to adopt them at scale. The analyst community’s reporting on AI project abandonment and failure rates underscores the same lesson: pilots proliferate; durable, audited production deployments require governance, data readiness, and change management.Practical 90‑day checklist for firm leaders (turn predictions into measurable steps)
- Inventory: catalog current engagement binders, DMS systems, and which clients still use paper or ad‑hoc PDFs.
- Pilot selection: pick one vendor agent and one extraction/ingress tool; demand sample replays and event‑level logs.
- Contract controls: update vendor agreements to include SOC/ISO attestations, model‑update notifications, and indemnities for data misuse.
- Training sprint: roll out a four‑week learning program focused on prompt hygiene, human‑in‑the‑loop verification, and AgentOps concepts.
- Security posture: engage an accounting‑aware MSSP for a 90‑day risk assessment and prioritized remediation roadmap.
- Pricing and SKU review: where automation reduces effort, model new fee schedules and advisory service packaging.
- Metrics: define and track KPIs — accuracy rates, time per engagement, exceptions per 100 documents, and client satisfaction.
Strengths, risks and final verdict
Strengths of the predictions- Grounded: the list ties vendor trends to firm operations and seasonality constraints.
- Actionable: most predictions map to discrete pilots and governance changes.
- Timely: the emphasis on agentic systems and Copilot in Fabric aligns with major vendor roadmaps.
- Some numeric claims (e.g., CPAFMA’s 66% OCR stat, exact percent increases in attacks) could not be independently verified in public sources and should be treated as directional until validated by your firm’s data or the original survey.
- Predictions that assume rapid behavior change (email abandonment, wholesale DMS replacement) understate the cultural and contractual inertia present in many firms.
- Vendor consolidation can be both an opportunity and a risk: cheaper acquisition of IP often accompanies integration and support challenges.
- Prioritize change management and training as much as software selection.
- Require reproducible pilot metrics and event‑level auditing before productionizing agentic tools.
- Treat Copilot and Power BI as a strategic analytics axis for 2026 pilots — but protect governance and provenance.
Kepczyk’s 2026 predictions present a pragmatic, vendor‑aware roadmap for accounting firms: adopt pilots where they reduce predictable seasonality pain, invest in people and governance to translate pilot wins into margin expansion, and plan for a technology landscape that will be dominated by agentic features delivered by established vendors and platform‑level copilots. Use the short checklist above to convert this foresight into a defensible, measurable plan for 2026 — one that protects professional responsibility while capturing the productivity upside of automation.
Note: claims that could not be independently verified from public sources (specific survey percentages and some headline attack‑rate figures) are explicitly flagged above and should be validated against primary surveys, vendor proof‑of‑value results, or your firm’s own telemetry before being used in policy changes or procurement decisions.
Source: CPA Practice Advisor Technology Predictions Impacting Your 2026 Accounting Firm Strategic Plans
