The memo.

There are two rooms that matter in this business, and almost nobody has stood in both.

The first room is the back office of a small plumbing company that a private equity firm just bought. The office manager has run the books for years. She knows that the "Other" account in the accounting system is where the warranty work goes, that the Tuesday invoice run has to happen before the bank sweep, and that the new owners keep sending her spreadsheets she does not have time to fill in. She has heard about AI. She has been pitched tools. She did not sign up for any of them, because none of them could tell her what would happen to her Tuesday.

The second room is the operating partner's office at the firm that bought her company. He has more portfolio companies than people to send, and is closing more this quarter. Every one of them runs on a different accounting system, invoices a different way, and has one person like her who holds the process together. His value-creation plan promised margin from operational improvement, and in a company that size operational improvement means the work she does every Tuesday. He has nobody he can send to change that work company by company, each one a different way.

Anaphora's founding insight is that these two people have the same problem and neither of them can solve it alone. The back-office work has to change, and nobody will do the work of changing it inside each business. She cannot adopt AI because nobody will fit it to her business. He cannot automate his portfolio's back offices because every company is a snowflake and his team is a handful of people.

We have been in both rooms. We understand why non-technical businesses hesitate, which is not fear of AI but fear of being handed a tool and left alone with it. And we understand what a PE firm is actually buying when it says "operational improvement," which is not software but certainty that the work will get done.

Everyone building workflow automation is fighting for the same customers: technical teams who can configure a tool. The businesses that cannot configure anything are unreachable to them, and they know it. Their pipelines are full of demos to companies that will never finish onboarding.

Everyone building for private equity is building reporting. Dashboards, data rooms, portfolio-monitoring tools. They count the work after it happens. None of them do the work.

Anaphora sits at the intersection. The PE firm is the customer-acquisition channel that pure automation companies could never build, and the automation is the product that pure PE tooling could never deliver. Each half is a known-hard problem on its own. Together they solve each other's hardest part.

Two costs used to make small-business automation uneconomic: an integration for every odd system the business runs, and a person to write down how the work is actually done. Both are falling at once.

An agent can now operate the software a small company already has, from the computer the employee already uses: the desktop accounting package, the dispatch tool nobody has an API for, the carrier portal in a browser tab. And a recording of a person doing the work is enough to start: it shows the steps, the applications and the values that move between them, and it can be turned into a workflow an agent runs. Both halves exist in Anaphora today. The recorder turns an employee's session into draft workflows the employee reviews and publishes, and the Computer Use Agent runs a workflow on that employee's computer, checks the result against the workflow's success criteria, and stops for a person only at a sign-in or a permission prompt. Neither has yet carried a real company's work; that is what the first pilot is for.

They have better-known brands and, today, more polished tooling. They lose to us in this market for three reasons.

They cannot get in the door. A workflow tool needs a champion inside the business who will set it up. A rollup portfolio company does not have that person. Anaphora does not need one. We arrive with the acquisition and the mandate.

They do not own the outcome. They ship software; the customer is responsible for making it work. Anaphora is accountable for the workflows running and for the exceptions reaching a person. In a market where the buyer is a PE firm with a value-creation plan to deliver, accountability is the product.

They learn from one side. Their feedback loop is support tickets from technical users. Ours is the employee's corrections to what the recording showed, the operating partner's priorities, and the record of every run: what the agent did, where it stopped, what a person had to step in for. We learn what the work actually is and how to run it, at the same time.

Robotic process automation already drives screens. It needs a developer to script each process, and the script breaks when the screen changes. Anaphora starts from a recording and the goal, finds the controls on the live screen each time, and an independent check decides whether the run worked. That is what lets a deployment be recorded instead of built.

We do not compete with them, and we chose not to become one. What the firm sees in Anaphora is operational: for each company, which workflows exist, which versions are waiting on a decision, which runs are active and which stopped and why, with the recordings and run evidence behind them. Each company stands on its own. Anaphora does not roll companies up into a financial view or look for synergies between them; the firm's numbers stay in the tools it already uses for numbers.

Consulting firms and operating partners' internal teams do operational-improvement work today. They win the first deployment against us on relationships. They lose the tenth on cost, because their tenth deployment costs the same as their first and ours does not. Every Anaphora deployment leaves recorded workflows, workflow definitions an agent can run, and a run history. Theirs leaves a slide deck.

The moat is not any single agent or connector. It is the proprietary map of how small-business back-office work actually gets done, built from recorded workflows, the workflow definitions drafted from them, human corrections, and the outcome of every run, across every company we have deployed. Each company's data stays in its own workspace; what carries over is our knowledge of the work. By the hundredth company in a category, we deploy from a near-complete blueprint. Nobody else has that map, because nobody else was in the room when it was drawn.

The path is a sequence of three businesses, each one funding and de-risking the next.

Lower-middle-market PE firms close add-on acquisitions all year. Each one is a company whose back office runs on people, spreadsheets and habit, owned by a firm that needs it to run better. Taking that company's routine work onto agents is worth a real engagement, because the alternative is months of an operating partner's time and a value-creation plan that slips.

A firm doing several add-ons a year is a real account. A few dozen such firms is a real services business. Services revenue is how we fund the platform and, more importantly, how we buy the data that becomes the moat.

Shape: a services business whose margin improves every quarter as the workflow library grows.

Once a company's workflows run on Anaphora, they keep running. Agents carry the routine work, exceptions keep flowing to people, and new workflows get recorded as the business changes. That is recurring revenue per company, priced on the work the agents carry, and it stays through the hold period and past the exit.

A fund whose portfolio companies each pay a monthly platform fee for the agents running their back offices is a recurring account, with services on top for every new acquisition. Many lower-middle-market firms fit that profile.

Shape: recurring revenue first, services a minority of the mix, deployments handled by a small team of deployment engineers working from blueprints.

The rollups exit. The companies keep running on Anaphora under new owners. The people who ran them start new businesses and call us. The accountants and brokers who serviced them start recommending us. By this point Anaphora has automated the back office of many non-technical businesses without ever selling to one directly, and the workflow library covers every routine job in a small-business office.

This is where the automation market actually is: every business that has a back office and no way to automate it, a meaningful share of its back-office payroll, and no incumbent who has ever reached it.

Anyone can describe business three. The reason nobody has built it is that the first customers are unreachable. Business one solves that by borrowing the PE industry's acquisition machine. Business two turns each solved customer into a recurring relationship and a data asset. Business three is what those assets are worth once they compound. The wedge is small on purpose. The market it opens is not.

The product climbs one ladder, and each rung is a little more autonomy earned with evidence: an agent that reads the company's records and recordings, then read-only connectors to email and the accounting system, then shadowing an employee and comparing what it would have done with what they did, then scoped execution with review gates. Today the first rung and a bounded form of the last exist: the reviewers that read imports and recordings, and the Computer Use Agent that runs a workflow on an employee's computer. The connectors and shadowing between them are not built.

Year one. A handful of PE firms, a few dozen companies with workflows running on Anaphora, one category where we have done it enough times that the blueprint is real. The recorder, the agent runtime and the first read-only connectors are in production. The twentieth company deploys in a fraction of the time the first took, and we can show it in Anaphora hours per deployment.

Year three. More firms. Many more companies running on Anaphora. Services are a minority of revenue. A firm that signs a platform deal puts Anaphora in the hundred-day plan by default. The workflow library spans several categories, and cross-category jobs like collections and accounts payable deploy from a shared blueprint. We have the first companies that exited a rollup and stayed on Anaphora under a new owner.

Year five. Anaphora is the operating layer for a meaningful share of lower-middle-market private equity in the United States. Company after company. Each back-office role has an agent that knows only the role and finds the inefficiencies in it itself; people decide what it may do. The first inbound customers who were never acquired by anyone, referred by someone who watched it work. The workflow library is the most complete record of how small-business back-office work is actually done that has ever existed, and it is ours.

Beyond. The back office is no longer a department. It is a service, and Anaphora is the one delivering it.

Back-office work in every small and mid-sized business is going to be automated. The technology is arriving faster than the businesses can adopt it, and the gap is distribution and trust, not capability. Private equity rollups are the one place where a third party gets handed non-technical businesses with a mandate to change how they run. We are the team that understands both the hesitant business and the firm that bought it. We will use that position to put the back office of company after company on agents, one company at a time, build the map of how their work gets done, and turn that map into the platform every business runs its back office on. That is how a services wedge into a niche becomes one of the largest software companies of the decade.