Will AI companies capture more value selling outcomes than selling software tools?

Sarah Guo · Apr 28, 2023

A significant and underappreciated part of the AI opportunity is the services market, not the software market

AI does more of the work itself, so e.g. taking low-level legal work is a bigger pie than software sold to legal firms; that's also where the productivity gains and labor displacement come from

32:48 20VC: In AI Who Wins? Startups or Incumbents? What Happens to Wealth Inequality? Why Will $10BN+ Companies Only Have 10 People | Why Defensibility in Startups is BS & Speed is Everything? Why Large Groups Worsen Decision-Making with Sarah Guo

Miles Grimshaw · Sep 18, 2023

If AI is to be a genuinely transformative force, it will produce a new architecture and business model best described as selling the work rather than the software — moving from copilot to control center, with an SLA on work outcomes instead of on uptime

Most software today is a database with a form on top where users act as data-entry monkeys; AI lets the application do the work instead, so the contract shifts to guaranteeing outcomes (marketing efficiency, leads, books closed in two days) more like a BPO than like software

Scope: conditional on AI realizing its potential; very little of this exists so far; current copilot products are a sustaining rather than disruptive architecture

46:26 20VC: Benchmark General Partner, Miles Grimshaw on The Five Pillars of Venture Capital, Why Data Can Be a Trap When Early-Stage Investing, Investing Lessons from Missing Figma and Plaid & The New Business Model for AI & Why Co-Pilot is an Incumbent Strate

Harry Stebbings · Dec 20, 2023

In the AI era, software should be sold as the work performed rather than as seats

35:04 20VC Roundtable: Spotify, Adobe & Linkedin CPOs on How AI Changes The Future of Product, Why AI is Now the Product, How TikTok Changed Product, Why Cost is the Biggest Barrier to LLM Usage & Why Incumbents Can Adopt AI Faster Than Any Prior Innovation Cyc

Tomer Cohen · Dec 20, 2023

What companies are really buying when they 'hire' is task completion, not people, and AI vendors should price against that

We conflate hiring with hiring people, but the underlying purchase is completed tasks

35:09 20VC Roundtable: Spotify, Adobe & Linkedin CPOs on How AI Changes The Future of Product, Why AI is Now the Product, How TikTok Changed Product, Why Cost is the Biggest Barrier to LLM Usage & Why Incumbents Can Adopt AI Faster Than Any Prior Innovation Cyc

Scott Belsky · Dec 20, 2023

Hourly billing and function-by-function seat-based pricing are antiquated in the age of AI, and value-based business models are the exciting innovation vector

Huge swaths of work are paid by the hour (lawyers, designers) for what can be achieved in an hour, and AI collapses both the time required and the business functions that seats were counted against

35:19 20VC Roundtable: Spotify, Adobe & Linkedin CPOs on How AI Changes The Future of Product, Why AI is Now the Product, How TikTok Changed Product, Why Cost is the Biggest Barrier to LLM Usage & Why Incumbents Can Adopt AI Faster Than Any Prior Innovation Cyc

Sarah Tavel · May 6, 2024

Higher entry valuations are rational for AI companies that sell the work, because that market can be 10-50x bigger than selling software

Selling work prices against the cost of headcount rather than a 10% productivity improvement, and has an easier go-to-market because no employee has to adopt a new tool — which widens the market aperture and justifies a higher price

Scope: does not apply to companies selling productivity improvements or underlying technology

24:20 20VC: Benchmark's Sarah Tavel on Are Foundation Models Commoditising | Why Frontier Models Will Be Closed Source | Why the Value is in the Application Layer | The Future of AI is "Selling the Work" Not the Tools

Aravind Srinivas · Jun 5, 2024

Once models achieve true reasoning, the $20-a-month subscription business model breaks — people will pay large sums for a single session or single output.

The ROI on one genuinely superior reasoning output is enormous, like an investor paying for a two-year company forecast that would otherwise take months of work; paying even 1% of the ROI is worth it.

Scope: conditional on achieving artificial superintelligence / true reasoning

12:21 20VC: Perplexity's Aravind Srinivas on Will Foundation Models Commoditise, Diminishing Returns in Model Performance, OpenAI vs Anthropic: Who Wins & Why the Next Breakthrough in Model Performance will be in Reasoning

David Luan · Jun 24, 2024

In knowledge work the most valuable AI work will not be priced per unit of work or consumption, but on its ability to augment what the user can do

Price-per-work pricing assumes repetitive, commoditized, cookie-cutter tasks with no creativity, whereas AI agents give people leverage to do new and more creative things — and you pay a copilot or teammate for augmentation, not per task

Scope: consumption/price-per-work pricing will genuinely become true in some places

37:10 20VC: Why Foundation Model Performance is Not Diminishing But Models Are Commoditising, Why Nvidia Will Enter the Model Space and Models Will Enter the Chip Space & The Right Business Model for AI Software with David Luan, Co-Founder @ Adept

Zach Lloyd · Oct 17, 2025

Pricing AI coding tools per developer is the wrong frame; companies will pay orders of magnitude more than developer-tool budgets because the right question is what the software is worth to the business

Every company is now to some degree a software company and software is the biggest lever they have, so these are not productivity tools

Scope: hedged on the specific $10k/month figure ('maybe')

15:03 20VC: The Startup Adding $1M ARR Every Week | Competing Against OpenAI's Codex and Claude Code: Who Wins | Why Gemini is Failing and GPT-5 Is Winning | Do Margins Matter in a World of AI | The Ugly Truth About AI Coding with Zach Lloyd, Warp

Winston Weinberg · Jan 19, 2026

AI products can escape the 'hostage' dynamic entirely by aligning price to the ROI they generate, Palantir-style, rather than relying on switching costs

Law firm customers have won new mandates by building custom solutions in Harvey — a $1M product earning a $20M deal — and in-house teams save large amounts of money by saving time, so the value created is high enough to align vendor and customer

Scope: based on his observation of legal customers

37:04 20VC: How Model Performance is Plateauing | Two Key Rules for Effective Deal-Making | Company Building Lessons from Keith Rabois, Brian Halligan and Pat Grady | Why Enterprise AI Adoption is Years Off with Harvey CEO Winston Weinberg

Harry Stebbings · Aug 8, 2026

Companies will spend $100–200M on the output of a single model query because that one result is strategically decisive

For decisions like whether to sponsor the World Cup with a ten-year exclusivity, the answer is worth that much to a P&G, Coca-Cola, NVIDIA or Visa

Scope: only for the largest global corporations

64:11 20VC: The AI Boom Will Create Enormous Roadkill: Who Wins & Loses | Why Founders Should Never Take Multi-Stage Money at Seed | Why Triple, Triple, Double, Double is Good Enough

Harry Stebbings · Sep 27, 2023

The future of AI is selling the work rather than selling the software

50:04 20VC: "How Being a Founder Almost Killed Me"; We Have Lied to a Generation of Founders | The Hardest Truths About Being a Founder Revealed | Why AI Co-Pilot is BS, Seat Pricing is Over & User Interfaces are Stupid with Christian Lanng

Patrick Forquer · May 11, 2026

Legora's valuation only makes sense if you price it against the trillion-dollar legal services market rather than the roughly $40BN legal tech market.

Against the legal tech TAM the competitive set's valuations don't add up, but much of legal services is rote, repeatable document-extraction work that Legora can address long term.

Scope: $40BN legal tech figure varies depending on who you ask; the services opportunity is a long-term view

13:21 20VC: Inside Legora: $100M ARR in 18 Months | Jude Law Generated $50M in Sales Pipeline: The Economics Broken Down | Competing Against Harvey, the 800 Pound Gorilla | Why Legora is Undervalued at $5.5BN with Patrick Forquer, CRO @ Legora

Julien Bek · Aug 24, 2026

The next trillion dollar company will be a software company that masquerades as a service business — selling outcomes rather than tools

Models are now reaching human parity so agents can complete tasks end to end; there's roughly a $1 to $6 ratio between spend on tools and spend on the associated service, and the prize is capturing the $6

Scope: it must be a software company underneath, not an actual services company

58:14 20VC: Inside Sequoia's Investment Committee: Lessons from Don Valentine, Doug Leone and Alfred Lin | How the SpaceX and Citadel Deals Went Down | What Sequoia Specifically Looks for in Founders with Julien Bek

Harry Stebbings · Mar 10, 2025 · hedged

Outcomes-based pricing almost doesn't work when there is a human in the loop.

Scope: 'almost'

51:39 20VC: Lessons from Investing $2BN and Returning $8BN in Cash | Why Most Venture Partnerships are Broken | We Sold Salesforce Early and Lost Out on Billions | Are The Best Deals Always Expensive and Competitive with Jake Saper @ Emergence Capital

Harry Stebbings · Mar 18, 2025

Outcome-based pricing is impossible unless the human is entirely out of the loop

With a human involved there will always be a dispute over attribution — the customer will claim they did 80% of the sale and the agent only 20%, so they won't pay full price for the outcome

Scope: self-flagged as a bold statement unsupported by data

8:44 20VC: HubSpot CEO on Where Value Accrues in SaaS AI | How HubSpot Competes Against Salesforce | Why B2B Is Not a Winner Take All Market | How to Go From SMB to Enterprise an Win | How SEO Dying Changes HubSpot's Business with Yamini Rangan

Miles Grimshaw · Sep 18, 2023 · hedged

The new AI business model could unlock previously underserved market segments and catalyze consumption, but that is different from being a new means of distribution as the internet was

Analogous to how SaaS architecture allowed selling to SMBs and PLG, but AI doesn't literally give you a new channel to reach customers

68:31 Ai business model unlocks underserved segments though unlike a new distribution channel

Mike Cannon-Brookes · Oct 13, 2025 · hedged

Outcome-based pricing has two structural problems: buyer and seller must agree on an accurate measurement of the outcome, and savings-sharing only works for one cycle because the new lower cost becomes the baseline

In a support-ticket example, charging a share of the saving from $100 to $80 works in year one, but in year two the cost is already $90 so you cannot keep charging for the same discount

42:12 Outcome pricing breaks down because measurement is hard and savings baseline resets each cycle

Kieran Flanagan · Jul 11, 2025

AI agents will be priced against human labor costs and outcomes, not software budgets

15:15 Agents priced against labor cost not software budget

Michael Eisenberg · Jun 19, 2024

Consultancies like McKinsey are thriving on at-risk fee models where they promise top- and bottom-line improvement using their own software and data

The fastest growing part of McKinsey's business is software, and they combine smart people with technological capability to upgrade clients

19:22 Consultancies capture outcome based fees by combining people with owned software

Michael Eisenberg · Jun 19, 2024

Corporate buyers now want new revenue opportunities rather than incremental efficiency gains, and traditional SaaS software cannot deliver that

Squeezing the incremental dollar of efficiency is no longer what interests buyers; generating new revenue requires a big leap

20:35 Buyers want new revenue not efficiency so traditional saas pricing fails

Michael Eisenberg · Jun 19, 2024

Buyers today are willing to pay for value and for work delivered, and this outcome-based pricing will come into software very quickly

Consultants like Accenture and McKinsey are already at the leading edge of charging for business improvement rather than seats

21:18 Buyers already pay for value so outcome based pricing will soon arrive in software

Harry Stebbings · Feb 16, 2026

He will not invest in companies that sell per seat; labour displacement is the only investment thesis that can return enough money to make funds worth it

Fund sizes now require returns that only job-replacing companies can generate

20:53 Only labour replacing pricing clears the fund return bar

Harry Stebbings · Oct 6, 2023

The phase where humans still make the decisions and AI assists is transitory; the true transition is selling the completed work rather than the tools, with AI executing end to end without human decisions

Tasks like writing marketing copy or choosing which micro-influencer to spend on won't need a human decision — you'll just specify a target CAC/LTV and a budget and the system will do it

30:33 Current tool assistance phase is transitory toward fully autonomous outcome selling

Guillermo Rauch · Oct 6, 2023

Even if Harry is right over a long enough timeline, the most successful innovations meet the world where it is, so AI and traditional software will coevolve rather than one replacing the other

The iPhone won by forking macOS rather than building fresh, and its first killer app was shrinking existing websites; native capabilities only emerged later, and twenty years on we still live in a hybrid of old and new DNA

31:15 Ai and traditional software coevolve rather than full replacement

Harry Stebbings · Aug 24, 2026 · hedged

Outcome-based selling only really works in markets with clear, definable resolution — it breaks down in ambiguous domains like sales and marketing where attribution is unclear

In customer support a ticket is resolved or not, but in marketing you can't tell which touchpoint produced the outcome

60:33 Works only where outcomes are clearly attributable

Julien Bek · Aug 24, 2026

Copilot products that sit inside the human judgment loop can capture that judgment as training signal, so today's human judgment becomes tomorrow's machine intelligence

Models handle verifiable intelligence well but not taste and body-language-type judgment that wasn't in training data; products in the loop can harness it

60:53 Copilots in the judgment loop unlock ambiguous domains

Jake Saper · Mar 10, 2025

Pricing will move toward outcomes-based over time, but it is hard today and we are still early.

Causality is hard to establish when support tickets have multiple human and bot touches, and disputing attribution creates an antagonistic monthly negotiation with the customer instead of a simple bill.

50:53 Outcomes based pricing is the future but attribution difficulty makes it hard today

Jake Saper · Mar 10, 2025

The easiest form of outcomes-based pricing today is AI-enabled services, where the vendor owns the entire delivery rather than selling a tool.

When you deliver the whole outcome there is no ambiguity about whether you did it — as with Mechanical Orchard promising mainframe migration in half the time at 80% of incumbent cost, with no payment if it fails.

51:40 Ai enabled services owning full delivery is the easiest outcomes pricing model

Jake Saper · Mar 10, 2025

Whether AI-enabled services suffer margin degradation depends entirely on pricing: labour-based pricing puts the risk upfront and is effectively a bet on your own AI working.

If you price on labour and the AI underperforms, margins are crushed; if it works, you capture far more margin than a traditional services business.

52:42 Labor based pricing puts margin risk on the vendor while working ai captures more margin

Jesse Zhang · Sep 19, 2025

Application-layer companies get priced against the business problem they solve, while middle-of-the-stack infra companies get priced as model cost plus a margin

Buyers benchmark infra against underlying model costs, but benchmark applications against labor savings and revenue impact

19:21 Infra prices at model cost plus margin apps price at business value

Harry Stebbings · Mar 18, 2025

A horizontal product cannot apply the same outcome-based pricing across its customer base because the value of an outcome varies hugely between customer types

A high-ticket use case like raising from LPs is worth far more than a low-ticket customer like a florist

9:27 Horizontal products cant apply uniform outcome pricing due to value heterogeneity