The landscape · Consulting and accounting
Consulting and accounting: what AI is really doing
Brief me in 2 minutes
AI in Consulting and accounting.
The Big Four are not quietly piloting AI, they are publishing reports written by it, and getting caught: Deloitte refunded part of a $440,000 government contract and KPMG had to pull a flagship report after an outside audit found 40 of its 45 citations were fake. At the same time KPMG and EY are rebuilding the actual mechanics of an audit around AI that reads every transaction instead of a sample, and a wave of AI-native challengers is underpricing MBB-style consulting projects by an order of magnitude. The part that is real and not hype is the back-office: hundreds of executive assistant, recruiting and support jobs are already gone or offshored. The part still mostly promise is the billable hour actually dying, not just bending.
The game: we show you 6 things happening. You call each one real now, early or mostly hype. Then the receipts.
Shift 1 of 6
AI-written client reports become a recurring, public Big Four embarrassment
Real now
These are completed, documented incidents (a refund actually issued, a report actually pulled), not projections about what AI might someday get wrong.
How it works. Deloitte had to refund part of a $440,000 Australian government report after AI-fabricated citations were found in it, and months later an independent probe by GPTZero found that 40 of 45 citations in a flagship KPMG report on agentic AI were fake, with the report pulled from KPMG's own website. EY had already withdrawn a separate study for the same reason.
Overstated: This is not proof AI cannot help with research, it is proof that these particular firms shipped AI output with no human actually checking the citations, which every firm's own AI policy says should not happen.
Shift 2 of 6
Audit stops sampling and starts reading every transaction
Real now
This is described as already deployed inside live UK audits at KPMG and EY, not a pilot, and it is already producing side effects (fee pressure, junior training concerns) that only show up once something is actually in use.
How it works. KPMG and EY are expanding AI across UK audit work so systems screen the entire population of a client's transactions for fraud and errors, rather than the traditional method of sampling a subset and extrapolating. EY, which runs nearly twice as many audits per year as some rivals with 85,000 auditors globally, is using this to ease workload on audit teams, while KPMG's UK audit chief says it is changing what the auditor's job actually looks like day to day.
Overstated: Regulators are clear this does not shrink the firm's legal responsibility: it remains fully accountable for its audit opinion and still has to explain its reasoning on individual transactions, AI screening or not.
Shift 3 of 6
AI-native audit and advisory platforms move from add-on tool to the firm's own workflow
Real now
Fieldguide has paying firm customers and a completed $75M funding round behind it, and the Aprio partnership traces back to a 2023 pilot that has already expanded three times, so this is live deployment, not a demo.
How it works. Fieldguide, an AI-native platform built specifically for audit and advisory engagements, now runs role-specific agents across planning, fieldwork, review and financial-statement prep inside firms like Aprio, the 20th largest accounting firm in the US, and partners with data-analytics vendor MindBridge, which screens 100% of a client's transactions, so the two plug directly into each other's workflow instead of competing point tools. Aprio frames its own methodology as the thing shaping how the agents are built, not the other way around.
Overstated: Every one of these launches keeps a licensed human reviewer in the loop by design (Aprio's own leadership says the point is freeing staff for judgment, not removing them), so "agentic audit" still means agents that a professional signs off on, not autonomous audits.
Shift 4 of 6
AI-native challengers underprice MBB by an order of magnitude
Early
One seed-stage vendor with a live product and paying PE customers is a real wedge, but it is nowhere near proof that this pattern displaces MBB-scale due diligence work broadly yet.
How it works. DiligenceSquared, an AI-native platform, replaces the commercial due diligence work that private equity firms typically buy from McKinsey, BCG and Bain for $500K to $1M per project, using AI voice agents and automated synthesis with source traceability instead of a staffed project team. It raised a $5M seed round to expand among private equity firms who are its buyers.
Overstated: A $5M seed company replacing a $500K-$1M engagement is a claim about price, not yet proof of parity in judgment or accountability on a live deal; MBB's actual moat has always been partner relationships and blame-absorption on a bad deal, which a report generator does not replace.
Shift 5 of 6
Support staff, not senior consultants, are the first jobs actually cut
Real now
These are completed layoffs and offshoring moves with specific headcounts reported by Bloomberg, not a forecast of future job losses.
How it works. PwC's US arm laid off around 600 executive assistants, recruiters and other support staff this year, on top of roughly 170 recruiting staff cut in Buenos Aires, while KPMG Australia is moving about 200 executive-assistant jobs to the Philippines. Firms had already been offshoring these roles (EY to Argentina, McKinsey to Florida, Costa Rica and Poland) before AI arrived, and AI adoption is now used to justify cutting even the offshored versions of the same jobs.
Overstated: PwC describes this as "modernising and simplifying how we work," language that does not confirm or deny how much of the cut is AI versus plain cost-cutting, so treat the AI attribution as the firms' framing, not an independently audited cause.
Shift 6 of 6
Deloitte puts a number on how much of consulting AI actually eats, and it kills the billable hour
Early
The number is a forward-looking internal estimate, and the pricing experiments cited alongside it (Vaimo's value-based model) are being pitched to clients now, not yet the industry default.
How it works. Deloitte Australia calculated that AI will handle up to 30 per cent of its routine consulting tasks within three years, a figure the firm itself is treating as confirmation that the traditional billable hour cannot survive unchanged. Separately, digital consultancy Vaimo is pushing clients toward value-based pricing tied to business outcomes instead of hours logged, arguing that time spent has stopped being a reliable proxy for the value AI-assisted work delivers.
Overstated: This is a three-year internal projection from one firm's Australian arm, not a fee schedule that has already changed; most invoices industry-wide are still hourly right now, so read this as the argument for the shift, not proof the shift is complete.
The scary part
What happens to jobs in Consulting and accounting
What AI is already taking:
- Drafting first-pass compliance and research reports, including the citations and footnotes, before a human reviews them PCMag ↗
- Screening every transaction in a client's books for fraud and anomalies, work that used to mean manually sampling a subset of records Traders Union ↗
- Executive-assistant, recruiting and other back-office support work that used to be a stable six-figure career path Financial Express ↗
- Commercial due diligence research for private equity deals, previously a staffed engagement billed at consulting-firm day rates PR Newswire ↗
What stays human:
Being the licensed, accountable signature on the work stays human: a regulator can fine a named audit partner, sanction a named advisor or ban a firm from contracts, and no vendor in this briefing claims AI removes that liability. Client trust in a bad-news conversation, telling a board its numbers are wrong or its deal has a hole in it, also stays human, because that is the part every firm above still frames as its people's job, not the software's.
Get better · 15 minutes
The skill that keeps you valuable is verification: being the person who actually opens the AI-drafted report and checks every citation and number against the source before it goes out, which is precisely the step KPMG's own investigator said nobody at the firm had done.
Try this this week:
Take a report or memo you have already written by hand for a client. Ask Claude or ChatGPT to draft an executive summary of the same material, then spend 15 minutes checking every number and cited source against your original documents. Time how long real verification takes you; that is the number to plan your own AI workflow around, not the tool's speed claim.
Your move
The crowded lane is horizontal bookkeeping copilots for small accounting firms, where Soraban and a dozen similar tools already compete on the same unbillable-admin pitch. The open space Integral and DiligenceSquared are proving out is narrower and bolder: becoming the licensed firm yourself (a tax advisory firm, a due-diligence shop) instead of selling software to one, which captures the fee the incumbent used to keep.
Watch whether a startup's product replaces licensed, sign-off work (Integral's tax filings, DiligenceSquared's due-diligence reports) or just speeds up drafting that a human still redoes; the first category captures fees the incumbent used to keep, the second competes for seat licenses inside firms that can build the same copilot internally. The red flag to probe in a pitch: ask what happens the first time the AI's output is wrong in front of a regulator or an investment committee, and who is licensed to be blamed for it.
This quarter, run one existing client deliverable, a compliance report, a due-diligence memo, a first-draft audit workpaper, through an AI draft and then time how long real, line-by-line verification against source documents actually takes. That number, not a vendor's speed claim, is what should set your internal policy on when a partner has to personally check every citation before anything leaves the building.
The valuable skill is no longer producing a first draft fast, it is catching the one fabricated citation, missed adjustment or offshored blind spot before a client or regulator does, which is exactly the checking step KPMG's own investigator said nobody had done. First step: pick one AI-drafted report this quarter and personally verify every number and source against the original documents, and time it.
What's actually changing
- 1
AI-written client reports become a recurring, public Big Four embarrassment
How it works. Deloitte had to refund part of a $440,000 Australian government report after AI-fabricated citations were found in it, and months later an independent probe by GPTZero found that 40 of 45 citations in a flagship KPMG report on agentic AI were fake, with the report pulled from KPMG's own website. EY had already withdrawn a separate study for the same reason. The mechanism is the same each time: a generative model drafts research with confident-sounding footnotes, and nobody on the review chain checks whether the underlying source exists before it goes out under a Big Four letterhead.
Why it matters. These are not obscure internal memos, they are paid deliverables and public thought-leadership from firms whose entire product is supposed to be trustworthy judgment, so each retraction is a direct hit to the thing clients are actually paying for.
What's overstatedThis is not proof AI cannot help with research, it is proof that these particular firms shipped AI output with no human actually checking the citations, which every firm's own AI policy says should not happen.
DeloitteKPMGEYGPTZero
3 receipts
- The Australian government ordered Deloitte to refund part of the $440,000 it was paid for a compliance report after AI-generated errors, mostly fake citations, were found in it.
The Australian government recently ordered Deloitte to refund a portion of the $440,000 it was paid for a compliance report after discovering several errors
PCMag, 6 Oct 2025 ↗ - An investigation found 40 of the 45 citations in KPMG's flagship agentic AI report were fake, and the report has since been removed from KPMG's site.
GPTZero, an AI detection software, found that 40 out of 45 citation titles are fake.
City A.M., 11 Sept 2026 ↗ - GPTZero's CEO said the pattern is repeating firm to firm: KPMG this month, EY the month before.
Today it was KPMG, last month it was EY, and we can only expect the problem to get worse in the short term.
City A.M., 11 Sept 2026 ↗
- The Australian government ordered Deloitte to refund part of the $440,000 it was paid for a compliance report after AI-generated errors, mostly fake citations, were found in it.
- 2
Audit stops sampling and starts reading every transaction
How it works. KPMG and EY are expanding AI across UK audit work so systems screen the entire population of a client's transactions for fraud and errors, rather than the traditional method of sampling a subset and extrapolating. EY, which runs nearly twice as many audits per year as some rivals with 85,000 auditors globally, is using this to ease workload on audit teams, while KPMG's UK audit chief says it is changing what the auditor's job actually looks like day to day.
Why it matters. Full-population screening is a genuinely different audit than a sampled one, and clients and regulators will eventually expect it as the new baseline, which raises the bar for firms and vendors who are still selling sample-based methodology.
What's overstatedRegulators are clear this does not shrink the firm's legal responsibility: it remains fully accountable for its audit opinion and still has to explain its reasoning on individual transactions, AI screening or not.
KPMGEYBDO UKGrant Thornton
3 receipts
- KPMG and EY are expanding AI to analyze entire transaction flows in UK audits instead of relying on traditional sampling.
KPMG and EY accelerate AI deployment in UK audits, enabling analysis of entire transaction flows and increasing coverage beyond traditional sampling methods.
Traders Union, 15 Sept 2026 ↗ - EY employs 85,000 auditors globally and runs nearly twice as many audits a year as comparable firms, and sees AI as relief for teams handling that volume.
EY, which employs 85,000 auditors globally and carries out nearly twice as many audits each year, also sees AI as a way to reduce pressure on teams handling large volumes of work.
Traders Union, 15 Sept 2026 ↗ - KPMG has pressed its own auditor, UK firm Grant Thornton, to lower its audit fees to reflect AI-driven cost savings.
KPMG has already pressed its own auditor, UK Grant Thornton, to reflect AI savings in lower fees.
Traders Union, 15 Sept 2026 ↗
- KPMG and EY are expanding AI to analyze entire transaction flows in UK audits instead of relying on traditional sampling.
- 3
AI-native audit and advisory platforms move from add-on tool to the firm's own workflow
How it works. Fieldguide, an AI-native platform built specifically for audit and advisory engagements, now runs role-specific agents across planning, fieldwork, review and financial-statement prep inside firms like Aprio, the 20th largest accounting firm in the US, and partners with data-analytics vendor MindBridge, which screens 100% of a client's transactions, so the two plug directly into each other's workflow instead of competing point tools. Aprio frames its own methodology as the thing shaping how the agents are built, not the other way around.
Why it matters. This is the clearest sign that mid-size firms, not just the Big Four, are re-platforming around agentic AI as core infrastructure, which narrows the technology gap between a firm well outside the Big Four and the Big Four faster than headcount or hourly rates ever could.
What's overstatedEvery one of these launches keeps a licensed human reviewer in the loop by design (Aprio's own leadership says the point is freeing staff for judgment, not removing them), so "agentic audit" still means agents that a professional signs off on, not autonomous audits.
FieldguideMindBridgeAprioBDO UK
4 receipts
- Fieldguide raised $75 million in a Series C round led by Growth Equity at Goldman Sachs Alternatives.
Fieldguide, an AI-native platform built for audit and advisory practices, has secured $75m in a Series C funding round led by Growth Equity at Goldman Sachs Alternatives.
International Accounting Bulletin, 3 Feb 2026 ↗ - MindBridge's technology analyzes 100% of a client's financial transactions and now plugs directly into Fieldguide's agentic workflow.
MindBridge analyzes 100% of financial transactions with explainable AI, surfacing risk earlier and providing the financial context practitioners need throughout the engagement
Business Wire, 20 Aug 2026 ↗ - Aprio, the 20th largest business advisory and accounting firm in the US, is co-building audit agents with Fieldguide as part of a $300 million, five-year AI investment.
the 20th largest business advisory and accounting firm in the U.S.
CPA Practice Advisor, 5 Aug 2026 ↗ - The Fieldguide collaboration is one expression of Aprio's $300 million, five-year investment in AI and automation.
The collaboration is one expression of Aprio's $300 million, five-year investment in AI and automation.
CPA Practice Advisor, 5 Aug 2026 ↗
- Fieldguide raised $75 million in a Series C round led by Growth Equity at Goldman Sachs Alternatives.
- 4
AI-native challengers underprice MBB by an order of magnitude
How it works. DiligenceSquared, an AI-native platform, replaces the commercial due diligence work that private equity firms typically buy from McKinsey, BCG and Bain for $500K to $1M per project, using AI voice agents and automated synthesis with source traceability instead of a staffed project team. It raised a $5M seed round to expand among private equity firms who are its buyers.
Why it matters. This is a direct price attack on the highest-margin, most senior end of consulting work, not just the junior research grunt work firms have already conceded to AI, and it is PE funds themselves, the clients, choosing to buy the cheaper AI-native version.
What's overstatedA $5M seed company replacing a $500K-$1M engagement is a claim about price, not yet proof of parity in judgment or accountability on a live deal; MBB's actual moat has always been partner relationships and blame-absorption on a bad deal, which a report generator does not replace.
DiligenceSquaredMcKinseyBCGBain
2 receipts
- DiligenceSquared replaces due diligence work private equity firms normally buy from McKinsey, BCG and Bain for $500K to $1M per project.
DiligenceSquared's platform replaces traditional consulting workflows that private equity funds typically purchase from leading consulting firms such as McKinsey, BCG and Bain for $500K to $1M per project.
PR Newswire, 6 Mar 2026 ↗ - DiligenceSquared raised a $5 million seed round led by RELENTLESS, with Y Combinator participating.
DiligenceSquared, an AI-native platform that automates commercial due diligence and market research for investment teams, today announced a $5 million seed round led by RELENTLESS, with participation from Y C ombinator.
PR Newswire, 6 Mar 2026 ↗
- DiligenceSquared replaces due diligence work private equity firms normally buy from McKinsey, BCG and Bain for $500K to $1M per project.
- 5
Support staff, not senior consultants, are the first jobs actually cut
How it works. PwC's US arm laid off around 600 executive assistants, recruiters and other support staff this year, on top of roughly 170 recruiting staff cut in Buenos Aires, while KPMG Australia is moving about 200 executive-assistant jobs to the Philippines. Firms had already been offshoring these roles (EY to Argentina, McKinsey to Florida, Costa Rica and Poland) before AI arrived, and AI adoption is now used to justify cutting even the offshored versions of the same jobs.
Why it matters. The people losing jobs right now are not the billable consultants AI supposedly threatens most, they are the six-figure support staff who keep partners and senior consultants productive, which is the opposite of where most AI-and-jobs commentary points its attention.
What's overstatedPwC describes this as "modernising and simplifying how we work," language that does not confirm or deny how much of the cut is AI versus plain cost-cutting, so treat the AI attribution as the firms' framing, not an independently audited cause.
PwCKPMGEYMcKinsey
3 receipts
- PwC's US arm laid off around 600 executive assistants, recruiters and other support staff this year.
PwC's US arm alone laid off around 600 executive assistants, recruiters and support staff earlier this year after months of restructuring uncertainty among non-client-facing teams.
Financial Express, 21 May 2026 ↗ - KPMG Australia is separately moving about 200 executive-assistant jobs to the Philippines.
KPMG Australia is reportedly moving about 200 EA jobs to the Philippines.
Financial Express, 21 May 2026 ↗ - PwC also cut around 170 recruiting staff in Buenos Aires as part of the same restructuring.
PwC's layoffs included around 170 recruiting staff in Buenos Aires alongside roles in Tampa.
Financial Express, 21 May 2026 ↗
- PwC's US arm laid off around 600 executive assistants, recruiters and other support staff this year.
- 6
Deloitte puts a number on how much of consulting AI actually eats, and it kills the billable hour
How it works. Deloitte Australia calculated that AI will handle up to 30 per cent of its routine consulting tasks within three years, a figure the firm itself is treating as confirmation that the traditional billable hour cannot survive unchanged. Separately, digital consultancy Vaimo is pushing clients toward value-based pricing tied to business outcomes instead of hours logged, arguing that time spent has stopped being a reliable proxy for the value AI-assisted work delivers.
Why it matters. A firm the size of Deloitte putting its own number on the share of work AI will absorb is a different kind of signal than a vendor's marketing claim, because it is the seller of billable hours admitting its own core product is shrinking.
What's overstatedThis is a three-year internal projection from one firm's Australian arm, not a fee schedule that has already changed; most invoices industry-wide are still hourly right now, so read this as the argument for the shift, not proof the shift is complete.
DeloitteVaimo
2 receipts
- Deloitte Australia's own calculation is that AI will handle up to 30 per cent of its routine consulting tasks within three years.
Deloitte Australia's calculation that AI will handle up to 30 per cent of its routine consulting tasks within three years confirms the death of the traditional billable hour.
Australian Financial Review, 7 May 2026 ↗ - Consultancy Vaimo is moving clients to value-based pricing tied to outcomes because time spent no longer reliably tracks the value AI-assisted work delivers.
Vaimo's value-based pricing model shifts the focus from hours and task estimates to clearly defined business objectives and success metrics.
PR Newswire, 10 Feb 2026 ↗
- Deloitte Australia's own calculation is that AI will handle up to 30 per cent of its routine consulting tasks within three years.
Who's doing it
8 companies to know, 7 with a round we can source. Newest first where dated.
Search these companies in the startup database ↗ · Explore all research tools
IntegralAI-native licensed tax advisory firm
Operates as its own licensed German tax advisory firm, Integral Tax GmbH, where AI agents do invoice reconciliation, transaction posting, payroll and tax filing and a licensed professional reviews and signs off.
Latest · 16 Sept 2026Raised an €18 million Series A led by Mosaic Ventures and Reid Hoffman to expand its AI agents for bookkeeping, payroll and tax work. Tech Funding News ↗
€18M Series A, 16 Sept 2026 source ↗
AdaptiveAI-native project accounting for construction
AI agents that run construction project accounting workflows (job costing, billing, WIP reporting, compliance tracking) for mid-market general contractors and specialty trades.
Latest · 4 Sept 2026Launched Project Accounting Agents, specialized AI roles covering work-in-progress reporting, accounts payable, change orders, billing and compliance tracking. For Construction Pros ↗
$30M Series B, 17 Sept 2026 source ↗
RilletAI-native ERP and accounting automation
AI-native ERP and bookkeeping platform built to automate the finance department's close, reporting and accounting processes for venture-backed companies.
Latest · 23 Aug 2026Reached a $1 billion valuation on a $100 million Series C, and struck a distribution partnership with RSM US to support AI-driven finance modernization. Crowdfund Insider ↗
$100M Series C, 19 Aug 2026 source ↗
FieldguideAI-native audit and advisory platform
Agentic platform used by accounting and advisory firms to run planning, fieldwork, review and disclosure work inside one system, with practitioners signing off at each step.
Latest · 12 Aug 2026Launched Field Financials, a set of agents that draft, tie out and finalize audit workpapers by referencing prior-year reports against current trial-balance data. GlobeNewswire ↗
$75M Series C, 3 Feb 2026 source ↗
MindBridgeAI transaction analytics for audit
Analyzes 100% of a client's financial transactions with explainable AI to flag risk, replacing traditional sample-based audit testing.
Latest · 29 June 2026Partnered with BDO UK to embed its full-population transaction analytics across BDO's audit practice ahead of upcoming year-end audits. Business Wire ↗
DiligenceSquaredAI-native commercial due diligence for private equity
AI-native platform that automates commercial due diligence and market research for private equity investment teams, undercutting traditional consulting-firm engagement fees.
Latest · 6 Mar 2026Closed a $5 million seed round led by RELENTLESS, with Y Combinator participating, to expand among private equity firms. PR Newswire ↗
$5M seed, 6 Mar 2026 source ↗
NumericAI accounting automation for enterprise finance teams
AI platform that started in month-end close management and expanded into a broader finance data platform for enterprise accounting teams.
Latest · 19 Nov 2025Used its new funding to expand from close management into a broader finance data platform, following prominent operator investors joining the round. PR Newswire ↗
$51M Series B, 19 Nov 2025 source ↗
SorabanAI admin copilot for accounting firms
AI copilot targeting the large share of an accounting firm's workweek that is unbillable client admin, rather than the tax preparation work itself.
Latest · 12 Aug 2025Closed a Series A led by Altos Ventures to help accounting firms automate unbillable admin work rather than the tax work itself. PR Newswire ↗
undisclosed Series A, 12 Aug 2025 source ↗
Your move
Our read, built on the receipts above.
The crowded lane is horizontal bookkeeping copilots for small accounting firms, where Soraban and a dozen similar tools already compete on the same unbillable-admin pitch. The open space Integral and DiligenceSquared are proving out is narrower and bolder: becoming the licensed firm yourself (a tax advisory firm, a due-diligence shop) instead of selling software to one, which captures the fee the incumbent used to keep.
Watch whether a startup's product replaces licensed, sign-off work (Integral's tax filings, DiligenceSquared's due-diligence reports) or just speeds up drafting that a human still redoes; the first category captures fees the incumbent used to keep, the second competes for seat licenses inside firms that can build the same copilot internally. The red flag to probe in a pitch: ask what happens the first time the AI's output is wrong in front of a regulator or an investment committee, and who is licensed to be blamed for it.
This quarter, run one existing client deliverable, a compliance report, a due-diligence memo, a first-draft audit workpaper, through an AI draft and then time how long real, line-by-line verification against source documents actually takes. That number, not a vendor's speed claim, is what should set your internal policy on when a partner has to personally check every citation before anything leaves the building.
The valuable skill is no longer producing a first draft fast, it is catching the one fabricated citation, missed adjustment or offshored blind spot before a client or regulator does, which is exactly the checking step KPMG's own investigator said nobody had done. First step: pick one AI-drafted report this quarter and personally verify every number and source against the original documents, and time it.
Other industries
- FinanceBanks move agentic AI from chatbots to running back-office and dealmaking work
- Software engineeringSpaceX buys the leading AI coding startup outright, for $60 billion
- HealthcareAmbient scribes go from pilot project to clinical infrastructure
- LegalLegal AI vendors go from research assistant to agent that produces the deliverable
- Real estateLeasing and resident calls move to voice AI agents that never sleep
- Sales and marketingCustomer support agents now get paid per resolved ticket, not per seat
- Jobs, HR and recruitingAI runs the first-round interview, not just the resume filter
- InsuranceClaims move from adjuster-first to AI-first, humans supervise the exceptions
- CybersecurityAutonomous AI agents take over penetration testing and vulnerability hunting
- Retail and e-commerceShopping agents get a checkout button, and Amazon spends 2026 deciding who's allowed to press it
- EducationNYC and LA ban AI for students, not for teachers
- Media and creativeAI video splits into model labs that train and aggregators that resell
- Manufacturing and roboticsHumanoid robots move from demo livestreams to paid warehouse shifts
- Transport and robotaxisRobotaxis go from demo to metered fare across a dozen-plus US cities
- Energy and data-centre powerGrid rules get rewritten mid-buildout, state by state
- GovernmentGSA turns AI adoption into a government-wide subscription, then flips 'free' to metered
- Agriculture and foodAutonomous tractor fleets learn to run the tools behind them, not just the wheel