Copilot in Excel for Accountants: Inside Microsoft’s 25 New Finance Skills and What They Mean for Pakistan’s Finance Teams

Copilot in excel for accountants now drafts the work, not just the formula. On October 29, 2026, Microsoft shipped more than 25 new finance skills plus partner-built plugins covering reconciliations, journal entries, FP&A variance narratives, audit sampling and tax workpapers. The core skills are generally available to Microsoft 365 Copilot subscribers, while the heavier reconciliation and journal-entry automation sits behind Frontier early access. Every output still requires professional review before it touches the ledger.

Microsoft changed the job description of a spreadsheet on October 29, 2026, because the company stopped shipping Copilot as a formula explainer and started shipping it as a drafting engine for finance deliverables. The expansion adds more than 25 finance skills and partner plugins to Excel and the Copilot app, which means the output of an AI prompt is now a workpaper rather than an answer. Demand showed up before the launch: during a seven-day limited Insider availability window, over 8,000 unique users invoked the new accounting skills more than 44,000 times across 18,000 Copilot conversations. For Pakistan — where finance labour is both an export product through the BPO sector and a domestic cost centre inside fragmented ERP estates — the consequences land harder than in most markets. This analysis breaks down what shipped, what you can actually use today, what it costs, where it fails, and what local finance leads should do about it.

About this analysis: Written by The Verge PK’s technology desk, which covers global product launches in local terms — price, availability, licensing and what the rules here allow. Every figure, date and quotation below is sourced to Microsoft’s October 29, 2026 announcement or to reporting on it, and linked at the point of use.

Transparency note: The Verge PK has no commercial relationship with Microsoft, Intuit, Xero or any vendor named in this article. Availability tiers and licensing terms described here reflect what Microsoft published at launch; because tenant settings and regional billing differ, verify both in your own Microsoft 365 admin centre before committing a close cycle to any capability discussed.

What Copilot in Excel for Accountants Actually Does Now

Copilot in excel for accountants is a set of role-specific AI skills inside Excel and the Copilot app that drafts finance deliverables — reconciliations, journal entries, variance narratives, audit samples and tax workpapers — from general ledger, subledger and bank data, then hands each output to a human for review and approval before posting.

What the skills produce that earlier Copilot could not:

• Completed workpapers with supporting documentation attached, not just formula explanations
• Proposed reconciling items flagged against general ledger, subledger and bank statement data
• Balanced journal entry drafts that validate debits against credits before submission

The significance of the October 29, 2026 release is the change in the unit of output. Earlier Copilot answered questions about a model; the finance skills assemble the artefact a reviewer signs. Industry analysts at Ledger Lowdown framed the shift directly: “This is bigger than a chatbot that explains a formula. Copilot can pull together files, conversations, models, and data, draft the work, and leave a professional to decide what survives review. Copilot is moving from answers to workpapers.” That sentence is the frame for everything below, because a tool that produces workpapers inherits the control requirements that attach to workpapers.

The pre-launch testing figures indicate pull rather than vendor push. Brian Jones, Corporate Vice President of the Excel Product Group at Microsoft, disclosed the seven-day Insider numbers in Microsoft’s announcement, and the ratio inside them is the interesting part: roughly two to three skill invocations per conversation suggests accountants ran the skills repeatedly within single workpaper sessions rather than trying them once out of curiosity.

Metric	Figure	Window
Unique users testing the new finance skills	8,000+	7-day Insider availability
Finance skill invocations	44,000+	Same window
Copilot conversations containing a finance skill	18,000	Same window

Inside the 25 Finance Skills: Five Categories, One Goal

Microsoft targeted what it describes as the repetitive middle of finance work, because that is precisely where hours concentrate and professional judgement does not. The skills cluster into five functional categories. The table below maps representative skills rather than a numbered inventory of all 25, since Microsoft announced “more than 25” capabilities plus partner plugins without publishing a one-to-one list.

Category	Representative skills	Output artefact
Bookkeeping and close	Balance comparison, discrepancy flagging	Proposed reconciling items
Journal entries	Accruals, reclassifications, allocations, reversals, recurring	Balanced entry draft
FP&A	Plan-versus-actual, driver decomposition	Variance narrative report
Audit and compliance	Sampling, control testing, traceback	Test log and exception register
Tax	Supporting calculations, payment estimates	Tax workpaper and filing calendar

Bookkeeping and Month-End Close

Copilot compares account balances and transactions across the general ledger, subledgers and bank statements to surface discrepancies, flags missing records, and prepares proposed reconciling items with supporting documentation attached for human review. The mechanic matters less than the consequence. Because the three-way comparison step is the single largest time sink in most closes — and because it is mechanical matching rather than judgement — automating it compresses close-cycle days rather than headcount. This is where copilot excel reconciliation changes the shape of the month: the team stops hunting for the break and starts arguing about whether the proposed treatment is right. Copilot month end close automation therefore reallocates effort toward exception resolution, which is the part that actually requires a qualified accountant.

Journal Entries: Accruals, Reclassifications, Reversals

Copilot journal entries in Excel cover the full routine set, and the engine checks for required inputs and validates that debits equal credits before submission.

• Accruals for expenses incurred but not yet invoiced
• Reclassifications between accounts or cost centres
• Allocations across departments, entities or projects
• Reversals of prior-period provisional entries
• Recurring entries with validated required inputs

The risk here deserves stating plainly: an engine that balances an entry has not established that the entry is correct in substance. A perfectly balanced accrual built on the wrong estimate posts cleanly and misstates the period. Consequently, approval workflow becomes more important once drafting is automated, not less, because the arithmetic check that used to catch sloppy manual entries no longer functions as a quality signal.

FP&A: Variance Analysis and Narrative Reporting

Copilot analyses financial performance against plan, decomposes variance by price, volume, mix, rate and timing, and drafts the narrative explaining what changed and why. The commentary layer that historically consumed analyst days arrives as a first draft within a session. As a result, analyst value shifts from writing the attribution to challenging it — asking whether the driver Copilot assigned to volume was actually a timing effect, or whether a favourable mix variance is masking a pricing concession. The draft is cheap; the challenge is where the expertise now sits.

Audit and Compliance: Sampling, Testing, Traceback

The audit skills assist with engagement planning, sample selection, control testing and tracing transactions back to source records, and they log test results, exceptions and corrective-action tracking. Copilot audit testing with AI introduces a documentation obligation that did not exist with manual selection, because audit evidence produced by a model needs its own provenance trail. Reviewers, engagement partners and regulators will ask how the sample was selected, from which population, and on what date the data was read. If that chain is not recorded at the moment of generation, reconstructing it later is effectively impossible — which means the provenance log has to be part of the workflow rather than an afterthought.

Tax: Workpapers, Estimates and Filing Calendars

Copilot tax workpapers cover supporting calculations for planning and compliance, payment estimates, and tracking of filing requirements and due dates. The jurisdictional limit is worth stating honestly: these skills were built around the rule sets Microsoft and its launch partners support, which are weighted toward the US, UK and other major markets. For Pakistani practitioners, that means FBR-specific treatments, withholding schedules and provincial sales tax regimes require local configuration and manual verification. Using the skill to assemble the workpaper structure is reasonable; trusting it to apply a Pakistani withholding rate without checking is not.

Cowork, Frontier and General Availability: What You Can Use Today

Microsoft split this release across two release tiers, and the logic behind the split is defensible. The core finance skills shipped to General Availability because they draft, summarise and explain — a wrong variance narrative embarrasses an analyst but does not misstate a balance sheet. Reconciliation and automated journal entry drafting sit in Frontier, Microsoft’s early-access programme for experimental features, because an error in those paths writes toward the books. Cowork is the third piece of vocabulary: it is the Copilot agent that chains multistep tasks across files, conversations, models and data sources, and it hosts the Finance and Capital Markets plugins.

Capability	Availability tier	Licence or setting needed
Core finance skills in Copilot in Excel	General Availability	Microsoft 365 Copilot
Capital Markets plugin in Cowork	General Availability	Microsoft 365 Copilot
Finance plugins in Cowork	General Availability	Microsoft 365 Copilot
Automated reconciliations	Frontier early access	Copilot licence plus tenant opt-in
Automated journal entry drafting	Frontier early access	Copilot licence plus tenant opt-in
Audit sampling and control testing	Frontier early access	Copilot licence plus tenant opt-in

Page not found – theverge.pk

The practical access question resolves at the tenant level, not the desktop. Because Frontier is opt-in, availability depends on what your administrator has enabled, which means a finance lead who plans a close around automated reconciliation without first raising it with IT will discover the gap during the close itself. The gate on everything described in this article is a Microsoft 365 Copilot licence — a premium per-user add-on rather than an included Microsoft 365 feature. Microsoft has not published Pakistan-specific pricing terms in this announcement, and because local purchases settle under regional billing arrangements that move with exchange rates, CFOs should price the licence from a current quote rather than from a headline US figure.

The Partner Plugin Layer: QuickBooks, Xero, Dynamics 365 and Market Data

Microsoft is distributing these integrations through its plugin registry rather than building them into Excel, which means coverage expands without an application update and without a customer-side deployment project. The announced launch partners span two distinct buyer groups.

Partner	Category	Finance use case
Microsoft Dynamics 365	ERP	Ledger, close and reporting data
Intuit QuickBooks	SMB accounting	Bookkeeping and reconciliation data
Xero	SMB accounting	Bookkeeping and bank feed data
Zoho Books	SMB accounting	Invoicing and ledger data
Moody’s	Market and credit data	Credit and risk inputs
FactSet	Market data	Valuation and benchmarking inputs
LSEG	Market data	Pricing and reference data
S&P Global	Market data	Ratings and index inputs

Adoption follows the data, not the interface. Because QuickBooks and Xero dominate small and mid-market bookkeeping across the US, UK and Australia, their inclusion determines whether an outsourced accounting team in Karachi or Lahore can point Copilot at a live client ledger instead of a stale exported CSV. That distinction decides whether the reconciliation skill saves an hour or creates a new reconciliation between the export and the source. For firms whose entire client base runs on those two platforms, the plugin layer is the product.

theverge.pk – AI Governance and Data Standards

The market-data plugins serve a different buyer entirely. Moody’s, FactSet, LSEG and S&P Global feed treasury, capital markets and valuation workflows, which signals that Microsoft is assembling a finance data fabric rather than shipping a bookkeeping feature. That ambition carries a governance cost worth naming: a plugin registry concentrates interoperability decisions in one vendor’s hands, which is exactly the dependency that open standards in AI are designed to counterbalance. Teams building permanent workflows on these connectors are accepting that the connector roadmap belongs to Microsoft.

What It Means for Pakistan’s Finance Teams

Copilot in excel for accountants matters more in Pakistan than in most markets, because the country’s finance labour is simultaneously an export product sold to foreign clients and a domestic cost centre inside fragmented local ERP estates. Both sides of that equation move when the reconciliation step stops being manual.

The BPO and Outsourced Accounting Multiplier

Because Copilot now connects directly to QuickBooks and Xero through the plugin registry, Pakistan’s BPO and outsourced accounting firms capture the gain before local corporates do. Reconciliation, bank matching and variance commentary are not incidental to those contracts — they are the billable hours being sold to US, UK and Gulf clients. Outsourced accounting in Pakistan has grown on a labour-arbitrage model where a Karachi team delivers the same workpaper at a fraction of the onshore cost. Automating the mechanical portion raises client volume per accountant, which consequently pressures the hourly pricing model that the sector was built on. Firms that keep billing hours will watch their own efficiency gain eat their revenue; firms that reprice toward deliverables and turnaround time convert the same gain into margin. That repricing decision is the strategic question this release forces on Pakistani practice owners.

From Doers to Reviewers: The Skills Shift

The sharpest warning on this release came from analysts at Ledger Lowdown: “the fastest way to make an accounting mess is to automate a mistake at scale.” Take the decisive reading of what that implies for the AI impact on accounting jobs: junior roles do not disappear, they re-specify. The work moves toward exception handling, anomaly detection, data scoping and governance — skills that a data-entry-trained junior does not possess by default. Because Copilot drafts and a human decides, a team of fast typists becomes less valuable than a team of sceptical reviewers. Therefore training budgets have to move before headcount does. A firm that cuts junior staff on the strength of a licence purchase, without retraining the remaining team to interrogate drafted output, has bought a cost saving and received a capability gap.

Excel as the Bridge Across Fragmented Local ERPs

Mid-sized Pakistani enterprises typically run a mix of local ERP packages, heavily customised accounting software and legacy systems that were never designed to talk to each other, which is precisely why ERP fragmentation stretches Pakistani finance close cycles into the third week of the month. Because Excel is the one surface every one of those systems exports to, Copilot inherits the role of reconciliation hub without requiring an ERP migration — the cheapest integration is the one you already own. The discipline to add is repeatability: treating extraction as a defined, documented pipeline rather than a monthly scramble of ad hoc exports is the same principle that governs automated data analysis pipelines in research settings, and it determines whether Copilot reads complete data or partial data.

Licensing Cost Versus Close-Cycle Time Saved

The CFO decision here is arithmetic, not sentiment. A per-user Microsoft 365 Copilot licence priced in US dollars lands very differently against a Pakistani payroll than against a US one, which means the break-even point sits at a different number of recovered close-cycle days in Lahore than in Chicago. Run the calculation on your own loaded cost per finance FTE per close day before buying seats. On Copilot Excel availability in Pakistan, the honest position is this: licensing is available to Microsoft 365 tenants, but Frontier access, plugin coverage for your specific accounting platform and the treatment of local tax rules should be verified inside your own tenant rather than assumed from a launch announcement.

Governance Guardrails: Stopping Copilot From Automating a Mistake at Scale

An AI that drafts entries inherits the control environment around it, and inherits its weaknesses too. Because a drafting engine repeats whatever logic it was given, a single mis-specified accrual rule propagates across every period it touches — and that multiplication is exactly the scale problem that financial controls exist to prevent. The control vocabulary for this is already written down: the NIST AI Risk Management Framework, published by the US National Institute of Standards and Technology in January 2023, organises AI risk management around govern, map, measure and manage functions, and it maps cleanly onto the finance control cycle.

Risk	Driver	Control
Mis-specified entry repeated monthly	Recurring template error	Dual approval on first run
Wrong data population sampled	Unscoped source selection	Documented scope sign-off
Unverifiable audit evidence	No provenance log for AI steps	Retain prompt and source trail
Confidential ledger data exposure	Broad tenant file access	Least-privilege data permissions
Over-reliance on draft narratives	Plausible but unchecked attribution	Reviewer challenges variance drivers
Local tax rule mismatch	Non-local rule coverage	Manual FBR verification before filing

Sequence matters in the review itself, because each step gates the next: a review that begins at the output rather than at the data scope cannot detect a wrong-population error, since the arithmetic will be internally consistent on the wrong dataset. Teams formalising this should align it with a documented AI data governance framework and understand where AI versus traditional data governance diverges, because model behaviour introduces controls that static data policies do not cover.

5 Critical AI Governance Challenges in Multi-Institution Research Labs – theverge.pk

1. Confirm the data scope and period Copilot actually read
2. Verify the source systems and whether any feed was stale or partial
3. Test the logic on a known-answer sample before accepting the output
4. Check the entry against policy, not only against arithmetic balance
5. Record the prompt, sources and reviewer in the audit trail
6. Assign named approval before posting, with no bulk auto-post

Copilot in Excel Compared to the Alternatives

Clear up the naming confusion first, because it drives wrong purchases. Copilot in Excel is the in-app assistant that received the new finance skills. Microsoft Copilot for Finance was the earlier Dynamics-connected offering aimed at ERP-centric workflows. Cowork is the newer multistep agent layer in the Copilot app that chains tasks across files, conversations and data sources, and it hosts the Finance and Capital Markets plugins. Three different products, three different buying decisions.

Criterion	Copilot in Excel	Copilot for Finance / Cowork	ChatGPT	Power Query
Primary strength	Drafting workpapers in-sheet	Multistep tasks across sources	General reasoning and drafting	Deterministic data transformation
Reconciliation reliability	Draft quality; Frontier tier experimental	Draft quality; agent-chained	No ledger-aware controls	Repeatable, same result each run
Ledger and ERP data access	Via partner plugins	Via Cowork plugins and ERP links	Manual paste only	Direct connectors to sources
Audit trail and governance fit	Tenant-governed; log prompts manually	Tenant-governed; log prompts manually	Outside tenant governance	Query steps are self-documenting
Licence requirement	Microsoft 365 Copilot	Microsoft 365 Copilot	Separate consumer or business plan	Included in Excel

Power Query still wins on deterministic, repeatable reconciliation, because a query returns the identical result on every run and a generative draft does not carry that property. Therefore the sensible configuration for a production close is not a choice between them — it is Power Query for the mechanical match and Copilot for the exception explanation and the commentary that follows. Teams that replace a working query with a prompt are trading a guarantee for convenience.

General-purpose chatbots lose on data residency and context rather than on raw capability. Brian Jones made the distinction explicitly: “General-purpose AI can help, but finance requires AI that understands the work for each role, draws upon the business context behind it, and supports professional review and judgment.” That is why pasting a client ledger extract into a consumer chatbot is a governance failure before it is an accuracy one — the data has left the tenant, the access log, and the retention policy that your engagement letter promised, and no amount of correct output repairs that.

Where Copilot Still Falls Short, and What to Keep Doing Manually

State the boundary decisively: Copilot drafts, it does not attest. Because the heaviest accounting skills — automated reconciliation, journal entry drafting, audit sampling — remain in Frontier early access, they carry experimental-feature status, which means they are unsuitable as the sole control over a statutory close or a filed return. Understanding Copilot Excel limitations is therefore part of deploying it responsibly rather than an argument against deploying it.

The second gap is provenance. AI-generated workpapers require a documented chain from source data to final output before an auditor can rely on them, and that chain is the user’s responsibility rather than the tool’s. The same difficulty recurs wherever models produce artefacts that others must trust, which is why model provenance challenges have become a governance discipline in their own right.

• Frontier features are experimental — keep a parallel manual reconciliation for at least one close cycle
• No deterministic repeatability on drafted output — re-run and compare before relying on a result
• Limited coverage of Pakistan-specific tax and withholding rules — verify every filing figure against FBR requirements
• Provenance of AI-generated workpapers is not automatic — log prompts, sources and reviewers yourself
• Partner plugin coverage excludes most local Pakistani ERPs — expect export-based workflows until connectors exist

Teams unwilling to accept these constraints have a reasonable alternative: keep Power Query or the ERP’s native reconciliation module as the system of record, and use Copilot only for exception narratives and first-draft commentary where a wrong sentence costs an edit rather than a restatement.

Frequently Asked Questions

What is Copilot in Excel for accountants?

Copilot in excel for accountants is a set of finance-specific AI skills inside Excel that drafts reconciliations, journal entries, variance narratives, audit samples and tax workpapers from ledger, subledger and bank data. Microsoft shipped the expanded set on October 29, 2026 alongside partner plugins. Outputs are drafts, which means a qualified professional must review and approve them before anything posts to the books.

What are the 25 new finance skills Microsoft added to Copilot?

They span five categories: bookkeeping and month-end close, journal entries, FP&A variance analysis, audit and compliance testing, and tax workpapers. Microsoft described “more than 25” skills plus partner-built plugins rather than publishing a numbered inventory, so the categories are the reliable unit for planning. Each category produces a defined artefact — reconciling items, balanced entries, variance narratives, test logs or tax workpapers.

Is Copilot in Excel accurate enough for financial reconciliations?

Accurate enough to draft, not to approve. The reconciliation skills sit in Frontier early access, which carries experimental-feature status by design. Run a parallel manual reconciliation for at least one full close cycle and test the skill against a known-answer sample before relying on its output. Because generative drafts are not deterministic, repeatability has to be verified rather than assumed.

Do I need a Microsoft 365 Copilot license to use the new finance skills?

Yes. The core finance skills in Copilot in Excel and the Finance and Capital Markets plugins in Cowork are generally available to Microsoft 365 Copilot subscribers. The Frontier-tier capabilities — automated reconciliations, journal entry drafting and audit sampling — additionally require a tenant opt-in that only an administrator can enable, so the licence alone is not sufficient.

What is Microsoft Frontier and how do finance teams get access?

Frontier is Microsoft’s early-access programme for experimental Copilot features, and it hosts the heavier accounting skills including automated reconciliations and journal entry drafting. Access is a tenant-level setting rather than a user preference. Consequently, a finance lead must request enablement through IT or the Microsoft 365 administrator, and should confirm it is active before scheduling a close around those capabilities.

Which accounting platforms connect to Copilot through partner plugins?

Microsoft Dynamics 365, Intuit QuickBooks, Xero and Zoho Books on the accounting side, plus Moody’s, FactSet, LSEG and S&P Global for market and credit data. The plugins ship through Microsoft’s plugin registry, which means coverage expands without an Excel update. Most locally built Pakistani ERPs are not in the announced set, so export-based workflows remain necessary there.

Will Copilot in Excel replace bookkeepers and junior accountants?

It replaces tasks rather than roles, at least at this stage. Because Copilot drafts work and leaves professional judgement to a human, demand shifts toward exception handling, anomaly detection, data scoping and review. The AI impact on accounting jobs therefore depends on retraining: firms that cut junior headcount without building review capability convert a cost saving into a capability gap and a control weakness.

How does Copilot in Excel change month-end close for Pakistani finance teams?

It compresses the reconciliation stage, which is where fragmented local ERPs consume the most time. Because every local accounting system exports to Excel, Copilot becomes the reconciliation hub without an ERP migration, so the saving arrives without a capital project. Verify Frontier availability, plugin coverage for your platform and local tax rule handling inside your own tenant before relying on it.

What are the governance risks of letting AI draft journal entries?

The primary risk is a mis-specified entry repeating every period, because a drafting engine reproduces whatever logic it was given. Add wrong-population sampling, unverifiable audit evidence and over-broad data access. Mitigate with dual approval on first runs, documented data scope sign-off, retained prompt and source logs, and least-privilege permissions aligned to the NIST AI Risk Management Framework.

How many people tested Microsoft’s new Copilot accounting skills before launch?

More than 8,000 unique users, during a seven-day limited Insider availability window. Those users invoked the new accounting skills more than 44,000 times across 18,000 Copilot conversations. The figures are attributed to Brian Jones, Corporate Vice President of the Excel Product Group at Microsoft, and the invocation-per-conversation ratio indicates sustained use within single working sessions.

The Bottom Line for Finance Teams

The capability is real, and copilot in excel for accountants closes a genuine gap, because the reconciliation and commentary work it drafts is the portion of the close that consumed hours without requiring judgement. What the release does not do is reduce the control burden — it relocates it. As a result, the reviewer carries more responsibility than before, since the arithmetic self-check that used to expose careless work now passes automatically on a well-formed but wrong entry.

For a Pakistani finance lead, the concrete next action is narrow and cheap. Check whether your tenant has Frontier enabled, confirm which of your accounting platforms appear in the partner plugin registry, and then run exactly one close in parallel — Copilot drafting alongside your existing manual or Power Query process — before changing any process documentation. That single comparison tells you more about your own data quality than any vendor benchmark will, because the errors it surfaces are yours.

The firms that benefit most will be the ones that move training budgets before they move headcount, since the scarce skill after this release is scepticism applied to plausible output. For more on how these tools land locally, read more AI coverage and ongoing technology analysis on The Verge PK.

References

• Microsoft. “Expanding Microsoft Copilot with more than 25 new finance skills and partner-built plugins.” Announcement, October 29, 2026. https://www.microsoft.com
• Ledger Lowdown. Analysis of Microsoft’s Copilot finance skills release, including the “answers to workpapers” and “automate a mistake at scale” commentary, 2026. https://ledgerlowdown.com
• CFO Dive. Reporting on Microsoft’s Copilot finance capability expansion, 2026. https://www.cfodive.com
• CPA Practice Advisor. Coverage of Copilot finance skills for accounting, tax and audit workflows, 2026. https://www.cpapracticeadvisor.com
• National Institute of Standards and Technology. “AI Risk Management Framework (AI RMF 1.0),” January 2023. https://www.nist.gov/itl/ai-risk-management-framework
• The Verge PK. “What Are Open Standards in AI?” https://theverge.pk/what-are-open-standards-in-ai-guide
• The Verge PK. “How to Build a Robust AI Data Governance Framework: A 6-Step Guide.” https://theverge.pk/how-to-build-ai-data-governance-framework-guide
• The Verge PK. “AI vs. Traditional Data Governance.” https://theverge.pk/ai-vs-traditional-data-governance-guide
• The Verge PK. “5 Common Model Provenance Challenges in Multi-Institution AI Labs.” https://theverge.pk/model-provenance-challenges-multi-institution-ai-labs
• The Verge PK. “How to Build an Automated Data Analysis Pipeline for Physics Research.” https://theverge.pk/how-to-build-automated-data-analysis-pipeline-physics-research

Related Reading

• Home – theverge.pk