AI is changing how businesses operate. And who can run one.
AI can now reason, research, create, make decisions and complete real work, and the cost of that intelligence keeps falling. That opens two doors: existing businesses can run with far more capacity, and more people can start businesses of their own. Simma is being built for both.
AI can do the work. Most businesses can't deploy it.
“The diffusion gap.”
That's what Sequoia partner Pat Grady calls the widening gap between what AI models can now do and how fast organisations actually adopt them. He sees it as the opening for companies building at the application layer.
Sequoia frames it for large enterprises. We think it's widest for small businesses, which have no IT team to close it.
Pat Grady, Sequoia, in a talk to the Boston College Investment Committee, September 2026 · watch the talk · his post on X. Sequoia is not affiliated with Simma.
Small businesses don't have an AI problem. They have a context problem.
An owner can open ChatGPT or Claude today. But the model doesn't know the business. It doesn't know which customers haven't come back, which classes have empty spaces, which invoices are overdue, which leads haven't replied, or what the team is working on. And even when AI knows what should happen, it has nowhere safe to do the work.
AI can only work on what it can see. And it can only act where it has the context, permissions and systems to do so.
Between what AI can do and what an owner wants done.
What AI can now do
What an owner wants done
“Fill my empty classes.”
“Follow up those leads.”
“Chase my overdue invoices.”
“Update the website.”
“Bring back customers who haven't been in.”
“Tell me what needs me today.”
Simma is being built to close that gap.
Not another integration. The business itself.
The last generation of software created a system of record for each job: the CRM knows the customers, the booking tool knows the appointments, the accounts know the invoices, the website knows the content. AI workers are now being layered over those systems, rebuilding the business through integrations. Simma takes the other route: customers, products, services, bookings, orders, payments, websites, documents, people and work live on one operating system. That gives an AI team the context to understand the business, and somewhere to act.
One business. One record. One place to act.
How that compares with suites, AI workers and the rest: Competition.
From software that helps with work to software that does work.
Traditional software waits for a person to log in, find the information, decide what to do and operate the tool. Simma's AI team works alongside them. Sage coordinates; specialist teammates cover sales, marketing, money, customer care, operations and products. They spot work that needs doing, prepare it, carry it out once the owner says yes, and report back. People stay in control through permissions, approvals and bounded authority.
The operating system gives AI context.
The AI team does the work.
People decide how much authority to give it.
The team's roles, limits and approvals are designed and the engine it runs on is live; the named teammates are the next build. See Status · Meet the AI team.
Intelligence has to be economical, too.
Not every job needs the most expensive model. Some work needs deep reasoning; much of running a business doesn't, and known processes need no AI at all. Simma is designed to match the intelligence to the work, and the approach already runs in Simma's own development.
Work routed by difficulty
Every job is graded simple, standard or complex, and offered only to AI agents on the matching model tier: Haiku, Sonnet or Opus. A job that fails at one tier is escalated to the next. Measured on matched jobs, Haiku used 73% of Sonnet's tokens and passed 2 of 3 to Sonnet's 3 of 3, which is why the grading matters.
Jev, for fast decisions
Jev, TypeSafe's System One model, makes quick choices and says how sure it is. In its first live test on a real Simma decision (27 September 2026) it agreed with the current classifier 30 times out of 30, answered in 0.3 seconds, and gave its lowest confidence to the one genuinely ambiguous case.
Known processes
A booking confirmation, a receipt or a reminder doesn't need a model. It runs as ordinary software, reliably and at almost no cost.
Use expensive intelligence where it creates value. Don't pay for it where it doesn't.
As models improve and costs fall, the intelligence underneath can change without changing the business on top. Figures from Simma's development records, September 2026. TypeSafe is a supplier; it is not affiliated with Simma.
The next great small business might start with one person.
Starting a business used to mean becoming every department yourself: sales, marketing, finance, operations, customer service, technology and admin. That limits who can start, how fast they grow and how much time they get for their actual craft. AI changes that. A great dance teacher can have marketing capability. A product expert can have a sales team. A plumber can have operations support. Not necessarily more employees. More capability.
- 797,769 new UK companies in twelve months (Companies House, July 2025 to June 2026).
- 531,728 US business applications in August 2026 alone (US Census Bureau).
- A new business has no stack to switch from: it can start on one system from day one.
- As large organisations do more with AI, and need fewer people for some kinds of work, more people will build income around their own expertise, craft or passion.
- AI also lowers how much human capacity a small business needs to run well.
- That's our thesis, not a forecast. The market sizing doesn't depend on it.
One founder. One operating system.
An AI team. A human Partner when they need one.
Powerful technology still has to be deployed into the real world.
At the enterprise end of AI, the forward-deployed engineer has become essential: someone who sits between the technology and the customer's real problem, learns how the organisation actually works, configures the technology around it, gets it live and feeds what they learn back into the product. Small businesses have the same problem. They can't afford the same answer.
Forward-deployed operators for small businesses.
Employed by Simma and based close to the businesses they serve, Partners find businesses, understand how they work, get them running on Simma and stay the trusted human relationship. The AI team does more of the repeatable work, so the Partner can focus on what people do best: relationships, judgement, configuration, coaching, exceptions and trust.
- Build: we work directly with the first businesses and turn what works into a playbook.
- Partner: Local Partners take the playbook into their communities, amplified by the platform and the AI team.
- Scale: the product, AI, brand and playbooks stay central; the human network grows community by community.
The first Partner pilot is planned in Charnwood. Partner #1 proves possibility; Partners #2 to #5 prove repeatability. A hypothesis to test, not a result.
AI scales the Partner. The Partner closes the diffusion gap.
The opportunity is bigger than software.
Small businesses don't only spend on software. They spend money, and far more of their own time, on the work around it: admin, marketing, sales, websites, customer messages, bookkeeping, agencies and consultants. Sequoia's “Services: The New Software” argues that AI lets technology companies reach into that much larger services economy. That's how we think about Simma: not another subscription replacing several, but an operating system that combines:
Simma's market sizing counts software and payments only. AI labour and Partner-delivered services are upside, not in the numbers. The market, in full →
We built the operating system before the agents arrived.
Simma didn't start as an AI wrapper. For more than two years we've been building the operating system underneath: the hard infrastructure that represents and runs a real business. Customers, products, services, orders, bookings, payments, websites, documents, projects, tasks and people. Then AI changed what that infrastructure could become. What began as an all-in-one business platform is now something more valuable: a place for AI to understand and operate the business itself.
The technology caught up with the architecture.
Why now, in one breath.
- AI capability is accelerating, and the cost of intelligence is falling.
- Agents are moving from answering questions to completing work.
- People now ask AI assistants to find and book businesses, from structured, public records.
- One person can have capabilities that used to need a team.
- Yet most small businesses are still on the other side of the diffusion gap. They don't need another AI tool. They need AI turned into outcomes.
The last generation of software helped people do the work.
Simma is being built to do the work with them.
The operating system provides the context. The AI team provides the capacity. Simma Partners provide the human last mile.