Last updated October 18, 2026
AI-native services
Sell the finished work at services prices, then make AI do more of it every year.
In this model, the company doesn't sell software. It sells the finished work, such as a reviewed contract, a placed insurance policy or a closed set of fund books, and uses AI plus a smaller team of experts to deliver it. Emergence Capital, the venture firm, describes these companies as built around AI from the start rather than bought and automated, and selling into labor budgets instead of IT budgets1. The bet is that clients pay services prices while AI does a growing share of the work, so margins rise toward software over time2,7.
The client buys a result, not a tool
A software customer has to learn the tool, run it and own the outcome. Here the vendor does the work and is the single point of accountability for it1. So clients hire these companies the way they'd hire a law firm or broker, often through referrals and personal networks. Emergence Capital points to the size of the prize: the Big Four accounting firms alone earn over $200 billion a year1.
Prices are inherited from the old firms
These companies usually price the way the incumbent did, not the way software does6. Crosby, an AI-native law firm, charges a fixed fee of about $400 per contract instead of billing by the hour5. Insurance agencies are typically paid a commission on the premium, and fund administrators typically charge a share of the assets they look after. Harper, an AI-native commercial insurance agency, and Hanover Park, an AI-native fund administrator, sit in those industries, though neither source states their exact pricing3,4. So there are three common ways to pay: per job, commission and basis points. They're usually separate choices rather than steps a client moves between, because each company inherits the pricing its industry already uses.
Margins start like services and should end up closer to software
Sasha McKenzie, a deal lead at Wellington Access Ventures, the venture investing arm of the asset manager Wellington Management, notes that gross margins start lower than software and improve as more of the work moves into the software2. Omar Khashaba, who writes Shifting Moats, sketches a hypothetical path from about 30% gross margin in year one to over 60% by year three7. That's the automation loop on the map: each job shows which steps AI can take over, so people handle only the exceptions and revenue per employee rises. Harper says one human salesperson handles 20 to 30 deals a month, while its AI handles more than 1,000 customers a month3.
People still check the work
The human review step isn't a temporary crutch. It's part of what the client pays for. Crosby has a licensed attorney review every document5, and at Hanover Park, AI agents propose journal entries that human accountants verify4. The risk is that a smaller team catches fewer mistakes. TechCrunch's Connie Loizos cites research in which 40% of employees had received AI output that looked polished but lacked substance8. An error that reaches the client is the drop-off that matters most here.
Growth comes from more teams and more services
Once a client trusts you with one kind of work, more of its teams can send work, and more people means more jobs. The other path is adjacent services. Harper's ambition is to go beyond placing policies into areas like risk and compliance3, and Hanover Park grew its assets under administration from $1 billion to $15 billion in a year4. Clients leave by taking the work in-house, often with AI software of their own, or by going back to a traditional provider.
How it flows
- Clients arrive through founder networks, outbound, referrals, industry events and partners.
- They sign an engagement priced like a firm's, not like software. Some decide switching is too risky and stay with their old provider.
- Work comes in, such as contracts, policies or ledgers.
- AI drafts the work and a licensed expert checks it4,5. Errors that reach the client break trust.
- The client gets the finished work back, often faster than from a traditional firm. Harper says its tasks take one to two days instead of five to seven. If something is wrong, the AI-native firm is responsible for it, not the client1,3.
- Each job shows which steps AI can take over, so people handle fewer of them and margins rise2,7.
- The client pays the way its industry already does, per job, by commission or in basis points, and sends more work from more teams.
- Trusted clients add adjacent services. Others take the work in-house or switch providers.
Metrics that matter
| Metric | Definition | Why it matters here |
|---|---|---|
| AI share of the work | Share of the steps in each job done by AI rather than people. | The engine of the model. As it rises, the cost per job falls while the price stays at services levels2,7. |
| Gross margin | Revenue minus the cost of delivery (people and compute), as a share of revenue. | Starts closer to a services firm than to software, and should climb each year2,7. If it doesn't, you're running a services firm. |
| Revenue per employee | Annual revenue divided by headcount. | Shifting Moats calls it the north star metric7. It shows whether growth comes from software or from hiring. |
| Turnaround time | Time from receiving the work to delivering it. | Speed is the visible difference from the old firm. Harper says tasks take one to two days instead of five to seven3. |
| Error rate | Share of delivered jobs with a mistake the client finds. | One firm is accountable for the result1, so an error that gets through costs trust and can end the relationship. |
| Jobs per client | Number of jobs each client sends per month. | Shows whether the client trusts you with more of the work, and whether more teams are sending it. |
| Services per client | Number of distinct services each client buys. | Drives the expansion loop. The client already trusts you, so adjacent work is the cheapest growth. |
Examples
Sources
- Jake Saper, Lotti Siniscalco and Rishub Nahar, “Why AI-Native Services, and Why Now,” Emergence Capital, March 23, 2026.
- Sasha McKenzie, “AI-native services,” Wellington Management, September 21, 2026.
- Dominic-Madori Davis, “AI insurance brokerage Harper raises $45M Series A and seed,” TechCrunch, February 25, 2026.
- Pulse 2.0, “Hanover Park: $27 million raised for AI-native fund administration platform,” March 29, 2026.
- Sacra, Crosby company profile (pricing, funding and service model), 2026.
- Omar Elayat, “False Profits,” The Verticalist (Euclid Ventures), October 6, 2026.
- Omar Khashaba, “AI-Native Services,” Shifting Moats, January 4, 2026. The margin path is a hypothetical example, not company data.
- Connie Loizos, “The AI services transformation may be harder than VCs think,” TechCrunch, September 28, 2025. Mostly about buying and automating existing firms.