--- title: "AI SDR Maintenance: What It Costs After Launch" description: "It is permanently under development, and that is the point. What actually needs maintaining, what CRM integration costs, and how to stay portable." author: "Eduard Klein" date: 2026-09-03 updated: 2026-09-03 lang: en type: knowledge tags: ["ai sdr", "maintenance", "crm integration", "vendor lock-in", "b2b sales"] url: "https://eduardklein.com/knowledge/ai-sdr-maintenance-what-it-costs-after-launch/" site: "Eduard Klein" --- # What ongoing maintenance does a built AI SDR require? It is permanently under development, and that is the point rather than the problem. What breaks is not the code. What moves is your market and your product, and the fit between them and the agent. Budget for continuous adjustment by the person who owns the sales process, not for a maintenance contract with an engineer. I am [Eduard Klein](https://eduardklein.com/about/). I advise CEOs, founders and boards on **AI Business Strategy**, and my own system has been under continuous development since 2023. The architecture is in [the full AI SDR blueprint](https://eduardklein.com/blog/built-own-ai-sdr/), published 19 August 2026. This page is what happens after launch. ## What "permanently under development" actually means You want the system to fit your needs exactly, and your needs move, because markets move and products change. I watched mine move twice in the 12 months to September 2026. Previously none of that lived inside the software. You worked on the software from the outside, and a release date and a vendor roadmap sat between you and any change. For an AI SDR to make sense, all of it has to go into the agent system instead. That is the work. It was always the work of an SDR. The difference in 2026 is that you do it with the agent rather than to a piece of software, which is collaboration rather than operating a tool. Concretely, in my own system the recurring work is correction. Every time I overrule a judgement I built the loop to store that correction and feed it back in as an example. After 6 months of that in 2026 the system sits measurably closer to my judgement than it did in March, and the volume of correction per week has fallen rather than risen. Maintenance that compounds downward is a different shape from maintenance that accrues. ## What actually breaks Data sources break, and access breaks. In 3 years the agent logic itself has almost never been the thing that failed. An API changes. A source I attached in 2026 changes its response shape. A key rotates and something 3 steps downstream stops quietly. That is the boring, real maintenance, and it is the same maintenance any integrated system has had for 20 years, since long before Claude or GPT existed. The second category is drift in fit. Nothing is broken and the output is worse, because your ICP moved or your product shipped something the agent has never heard of. No error appears anywhere. I caught 1 of those in 2026 by reading output, which is the argument for staying in the loop long after you trust the thing. ## How does an AI SDR integrate with our CRM? Over MCP or the vendor's own API. Nobody needs Zapier for this in 2026. That is last-century technology. The agent connects itself given the API documentation, which is a change since 2023 that most integration budgets have not caught up with. The only real challenge left is key management, and in enterprise environments there are now decent connector tools that let Salesforce and HubSpot and the rest talk to everything. The more interesting version of the question is the one I answered by building my own store. The CRM is not a system the agent integrates with. It is the spine the agent lives on, and I built mine because my agents need full read and write on every company, every person, every email in and out, and every decision with the reason attached. I tested that against commercial CRMs first. You spend your life fighting an API designed for a human clicking a mouse, and paying a per-seat licence for a robot. ## The reason I own the schema I built Component 3 myself for a second reason, and it is the one that pays back after launch rather than during the build. I own the schema. So lawful basis, retention, deletion stop being a policy document in a shared drive and become 3 columns in a table with real dates in them. You should get your own legal advice, because I am not your lawyer. What I can say from running it is that when you own the store, "what do you have on me and why" stops being a project and becomes a query you run while the person is still on the phone. Under the EU AI Act and GDPR that difference is worth more than the 2 weeks the tables took to design. ## How do we avoid vendor lock-in and keep our data portable? Keep the rights to your data, and put the database structure in the contract too. Data alone is not portable. Data plus schema is. With both you migrate into your own cloud database later and carry on. Without the schema you get a CSV export and a 3-month project to work out what the columns meant, which I have watched a company pay for twice. Any decent AI agency will not try to lock you in regardless, because in 2026 this is a service rather than a product. That is a real shift and it is worth naming when you negotiate. You are not buying a seat. You are commissioning a system you will own, and the contract should read that way. ## What technical expertise is required to build a custom AI SDR? Programming knowledge, software concepts, project management, build management, versioning, plus compliance and security. That is 7 competencies and it is the honest list. The public debate offers you "no-code is enough" or "you need an engineer", and neither is right. What you need is somebody who knows the concepts, even if Claude Code writes most of the actual code. After launch the requirement changes shape. You no longer need somebody who can build it. You need somebody who can tell whether a change is safe. That is a smaller job than 8 months of building, and a permanent one. In an enterprise, the point where everyone starts building their own is the point where you need central infrastructure instead of 40 private agents. The honest threshold is not headcount. It is comprehension. The moment you can no longer manage the system because you no longer understand it, you need a professional on board. ## What this costs in practice Almost nothing in licence fees and a real amount of attention. There is no per-seat line. A $20 model subscription in 2026, maybe $200. A cloud database. A VPS. Against an enterprise paying $600,000 a year in Salesforce licences, the running cost is a rounding error, and it does not grow when you hire. What it costs instead is that 1 person has to keep owning it. If that person leaves and nobody replaced the comprehension, you have a system nobody can safely change, which is a worse position than the licence you were avoiding. That is the honest risk of the build side, and it is worth pricing before month 1. If you want somebody who has run this past the launch to sit on your side of it, that is what my [AI coaching](https://eduardklein.com/coaching/) work is. Whether ChatGPT and Google AI Overview can see your company at all when a buyer asks is a separate question: [AI strategic visibility](https://eduardklein.com/blog/ai-strategic-visibility/). ## What I still do not know I do not know what this looks like at 5 years. My oldest agent code is from 2023 and in 3 years I have rebuilt every part of it at least once. I do not yet know whether that is a property of an immature field that settles down, or the permanent condition of running systems on models that change under you twice a year. Ask me in 2028. I will have an answer by then and I am not going to guess at it now. What you signed up for before launch: [the tech stack I actually run](https://eduardklein.com/knowledge/ai-sdr-tech-stack-the-one-i-actually-run/), [whether to build your own AI SDR or buy one](https://eduardklein.com/knowledge/ai-sdr-build-or-buy-why-i-built/) and [how long an in-house AI SDR takes to build](https://eduardklein.com/knowledge/how-long-to-build-an-ai-sdr-mine-took-months/). ## About the author **Eduard Klein** is an AI Business Strategist for CEOs, founders & boards, and he learned what a system costs after the signature from the vendor side: software sales for SAP, for Microsoft, for Oracle and for the German Stock Exchange, where the licence was the easy conversation and the years afterwards were the expensive one. Years in key account management and inside sales, a sales manager role in the software industry, an SDR quota up to 3.5 million in revenue, and 170+ articles across 100+ publications including LeadSquared and Corefactors sit behind the ownership economics here ([the full list](https://eduardklein.com/media/)).