Where this goes.

A few years from now, a regional HVAC rollup owns branches across several states. Each branch was once a family business with its own way of invoicing, its own spreadsheet for scheduling, and one person who knew where everything was.

Today every back-office role at every branch has an agent beside it. The agent knows only the role: the dispatcher's, the bookkeeper's, the person who chases late invoices. It was not handed a process map. It watched the work, read the records, and found the waste itself. At one branch it noticed that warranty callbacks were keyed into the dispatch tool and again into the accounting system, cited the sessions where it saw it happen, and proposed a workflow to do it once. The bookkeeper corrected one step and approved it. It has run every day since, and the morning she had to step in, it was because the supplier portal asked for a new password.

Each branch's workflows are its own. The rollup does not force its branches into one process; it lets each one's agents carry that branch's work the way that branch does it.

The people who used to do that work are still there. They handle the customers who call angry, the vendor who shipped the wrong compressor, the job that needs a judgment call. The agents do the rest and hand every exception to a human with a full account of what they did and where they stopped.

Nobody at the rollup set any of this up. The PE firm that assembled the company handed each branch to Anaphora on the day the deal closed. Anaphora connected the systems, recorded how the work was actually done, built the agents, and stayed accountable for them running. That is the whole product: back-office work, delivered as an outcome, to businesses that would never have built it themselves.

Every category of routine office work in every small and mid-sized business is the addressable market. Rollups are how we get there first.

The first HVAC company Anaphora deployed took a deployment engineer months. The tenth took weeks. The fortieth took days, and most of that was waiting on system access.

What changed was not the engineer. It was the library.

Every deployment leaves behind system mappings, recorded workflows, the workflow definitions drafted from them, exception-handling rules, agent run histories, and the human corrections that taught the agents where they were wrong. By the fortieth HVAC company, Anaphora already knew that dispatchers batch-schedule on Thursday afternoons, that warranty callbacks get coded several different ways, and that the "Uncategorized" account in the accounting system is where the interesting problems hide. No company's data ever left its own workspace to teach that; the knowledge of how the work is done is what carried over.

The hundred-and-first company in a category starts from a near-complete blueprint. The deployment engineer has an employee record a session, confirms the blueprint fits, adjusts the part that is genuinely different, and turns it on. What was a services engagement is now a deployment. What was a deployment becomes a checkbox.

Then the library crosses categories. Accounts payable in a dental rollup looks a lot like accounts payable in a landscaping rollup. Collections is collections. The workflow library stops being organized by industry and starts being organized by job. At that point Anaphora is no longer a services firm with good tooling. It is a compounding, proprietary map of how small-business work actually gets done, and nobody else has one, because nobody else was in the room for every one of those deployments.

Lower-middle-market private equity owns a great many companies at any given time, and buys more every year. Each acquisition triggers the same question: who is going to make this company's back office run better, and how, when it runs on one person's memory and a desktop accounting package?

In this story, that question has a default answer. A firm signs a platform deal, and Anaphora is in the hundred-day plan next to the lawyers and the lenders. Acquisition day is Anaphora day. The firm's operating partner does not ask "who is going to fix this one's back office?" because the answer has been the same for the last several.

The economics that make this stick:

The firm's cost to put a new acquisition's back office on agents falls every year while its deal pace goes up.

The operating team sees, for every company, which workflows run, which are waiting on a decision, and which runs stopped and why, without asking anyone for an update.

Each company's work runs the way that company does it. Nothing has to be merged before anything gets better.

Anaphora becomes the layer that every portfolio company in the fund runs its back office through, and part of the story the next fund is raised on: a firm that can show its companies' routine work is carried by agents, with a record of every run, has an operating edge it can prove.

The ambition is not a big customer list. It is that "we run on Anaphora" is a line in the diligence deck, the way naming your CRM stopped being a differentiator and became a baseline.

Most small businesses want what AI is supposed to do for them and have no idea how to get it. They have been sold chatbots and dashboards. They have signed up for tools, not configured them, and cancelled. They do not have an IT person. They are not going to become one.

Every pure workflow-automation company has run into this wall. The technology works; the customer cannot be reached, cannot be onboarded, and cannot be kept.

Anaphora's path around the wall is that it never sells to these businesses cold. A PE firm acquires them and Anaphora arrives with the deal. The business does not evaluate a vendor; it gets handed an operating team. By the time the rollup is sold, the business has had its routine work carried by Anaphora agents for years and cannot imagine going back.

Then the rollup exits. The new owner, a strategic buyer or another fund, inherits a company that runs on Anaphora. The former owners start new businesses and want the same thing. The branch manager who watched it work leaves to run a competitor and calls Anaphora first. The accountant who serviced the rollup starts recommending it to every client with a scheduling problem.

The PE wedge is how Anaphora gets its first wave of non-technical businesses without ever running a self-serve funnel. Those businesses are the seed of a distribution channel that the pure-play automation companies would have paid anything to build. Anaphora uses one industry's acquisition machine to bootstrap distribution into the largest and hardest-to-reach market in the economy.

The last story is about how the economics close.

A company that runs on Anaphora has its back-office work recorded, written down as workflows an agent runs, and checked on every run, with a history of what happened each time and every exception a person handled. A buyer doing diligence on it finds no black boxes. The key-person risk that normally discounts a small acquisition, the office manager of many years whose departure would stop the invoices, is smaller because the work no longer lives only in her head.

Those companies trade at a premium. Not a huge one at first, then a larger one, as buyers learn that "runs on Anaphora" means the back office is a known quantity. Firms start acquiring companies specifically to put them on Anaphora, because the difference between a business that runs on one person's memory and one that runs on documented, verified workflows is now visible.

At that point Anaphora is not a cost line in the operating budget. It is part of how the firm creates value. Anaphora's pricing follows the work: per company, for the agents carrying its back office, not per seat. The firm, the portfolio company and Anaphora all want the same thing, more of the routine work running reliably, which is the only arrangement that has ever worked in this market.

The end state: Anaphora is how private capital turns businesses that run on a few people's memory into businesses that run on documented, verified workflows, and the market prices that transformation. Everything upstream, the recorder, the workflow library, the agent runtime, the deployment engineer, exists to make that transformation cheaper and more certain each time.