Before you buy AI features, count what you already own

Every tool in the stack now sells an AI add-on: your CRM, your helpdesk, your video calls, your spreadsheets. Usually per seat, per month. Some of these features are very good, and we use plenty of them. But the deal has a shape to it that is worth looking at clearly before you sign up for all of them, because the individual terms are reasonable and the pattern that emerges from them is not always what you want.

Considered one at a time, each add-on does something useful. Considered together, three things recur. A copy of your data leaves for someone else's platform, on terms you renegotiate annually rather than control directly. Each product reasons about your business through a narrow slice of it: the CRM knows about the pipeline and nothing else, the helpdesk knows about tickets and nothing else. And the insight one of them produces cannot easily be combined with anything else you own, because it lives inside that product's interface and its export options. After a dozen subscriptions, the honest summary is not that you have built an AI capability. It is that you are renting a dozen small ones, none of which can talk to each other, and all of which renew next quarter.

The alternative is older-fashioned, which is to put the engineering where the data is. One system, connected across your CRM, helpdesk, documents and calls, answering questions with all of that context at once, grounded in your own data with the audit trail attached. Because it sees the whole picture, it can make connections no single product can: the support theme that predicts the renewal risk, the competitor mention that explains the lost quarter. And because it is yours, it gets better as your data gets better, it extends in whatever direction you need, and nobody can reprice it or discontinue it as part of someone else's strategic pivot.

None of this means buying is wrong, and it would be strange to leave that impression. When the workload is genuinely commodity, nothing sensitive is involved, and the feature needs no integration with anything else to be useful, the add-on is usually the right answer, and building in that situation is expensive for little return. Building earns its keep when the workflow is genuinely yours, when the data is sensitive or differentiating, and when the value compounds as the system learns more of your business over time. In our experience, most teams that look carefully find they have more of the second kind of work than they first assumed, because the most valuable questions in a business tend to be precisely the ones that span systems and cannot be answered from any single product's narrow view.

That's the approach we take on every engagement, and it's why most of our systems are in daily use within weeks of the first conversation. If you'd like to talk about what's worth building in-house for your team, get in touch.

Author: Michael Wells