How we replaced a patchwork of agency tools with one CRM built with AI, covering leads, audits, actions, paid media and client reporting, and the lessons we learned building it.
Case Study: How We Built Our Own Agency CRM With AI
Most agencies run on a patchwork of tools: a CRM for leads, a spreadsheet for tasks, a crawler for audits, a reporting tool for clients, and a lot of copying and pasting in between. We did too.
This case study explains how we replaced that patchwork with a single agency operating system, built with AI, around the way we actually work. It also covers the lessons we learned the hard way, because they are the ones most useful to any business considering the same step.
The starting point
Our work spans the full client journey: winning new business, auditing websites, improving search and AI visibility, running paid media, and reporting back to clients every month. Each stage had its own tool, and none of them talked to each other. Client context lived in people's heads and inboxes.
Off-the-shelf CRMs could handle contacts and deals, but not the parts that make our work distinctive: generative engine optimisation audits, AI visibility reporting, and the link between an audit finding and the action that fixes it.
So we built our own.
What the CRM does
The CRM now covers the whole agency, from first enquiry to monthly report.
Winning new business. Enquiries from our website's audit and contact forms land in a Lead Engine, where they are scored and enriched. Before any discovery call, a Quote Assessment takes a prospect's web address and produces a scored assessment, a draft proposal and talking points. Target lists and email campaigns handle outbound prospecting.
Auditing and improving websites. Website Brain crawls a site and produces a technical SEO health score with the critical issues listed. The GEO Audit Engine assesses how ready a business is for AI search, scoring entity clarity, topical authority, structured data, E-E-A-T and a baseline of AI mentions. Site Assessments turns crawl exports into AI-generated recommendations, and a keyword tracker follows rankings month by month.
Turning findings into work. Every recommendation becomes an action linked to the client and the problem it solves. An actions board shows every open item across all clients, and each person has their own task list, so nothing falls through the cracks.
Paid media. An ads manager brings Google, Meta, LinkedIn and TikTok accounts into one place, with AI-generated optimisation suggestions each month. A dedicated tracker records competitor ads seen inside ChatGPT and scores clients for advertising readiness on the new platform.
Reporting to clients. The CRM feeds Thursday Intelligence, our client portal. Each client has a private link showing their search performance, rankings, GEO audit and AI visibility, which is now built on measured data from Microsoft Copilot, Google AI Overviews and ChatGPT.
Running the agency. A morning view shows every client's health, open tasks and anything at risk. An agency dashboard covers pipeline, team capacity and client health. An operations monitor watches the automations, and a governance area includes a kill cord that can stop automated activity instantly if something misbehaves.
How AI built it
The CRM was built with an AI app-building platform, with AI doing most of the engineering and people setting the direction, the rules and the approvals.
That combination is the point. AI made it possible for a small agency to build software that would once have needed a development team. People made sure it did the right things.
What we learned
Building with AI is fast. Building with AI well takes discipline. These are the lessons we would pass on to any business attempting the same.
AI will fill a gap with something plausible. When a data source fails, an AI builder may produce convincing numbers rather than an error. We now treat every result as a claim until it has been checked against the live database. A change is only finished when the data and the live screen both confirm it.
Estimates must never look like measurements. Our original AI visibility score was modelled, not measured. It looked precise, so clients read it as fact. We retired it and rebuilt the reporting around first-party data from the platforms themselves. Wherever data is missing, the system now says "Not available" rather than showing a zero.
Small, verified steps beat big rewrites. AI builders are reliable for screens, forms and data structures, and much less reliable for complex multi-step logic such as fetching data from other services. We build that kind of logic in the smallest possible steps and verify each one before moving on.
People stay in charge of anything a client sees. Internal changes can move quickly. Anything that reaches a client portal is built, checked in preview and approved by a person before it goes live. Deleting data and anything involving money always needs explicit sign-off.
Write the rules down. We keep a running record of the architecture, every decision and every lesson learned, and update it whenever something new comes to light. It is what lets the work continue safely from one session to the next.
What changed
The agency now runs from one system instead of a dozen. Client context sits in one place. Audit findings lead directly to actions, and actions lead directly to what the client sees in their portal. Routine work, such as monthly AI visibility reporting, has been reduced to a few clicks per client, and the time saved goes into the thinking clients pay us for.
Most importantly, we now recommend AI transformation from experience. We know where AI saves time, where it needs supervision, and where a person has to make the call, because we have done it on our own business first.
Considering the same for your business?
You do not need to build a CRM to benefit from AI transformation. The approach is the same at any scale: find where AI genuinely helps, build it in carefully, and keep people in charge of what matters.
Read more about our AI transformation service, see how we ran our own agency through AI transformation, or request a free GEO audit.