The agent strategy you will execute was decided by your last five years of SaaS contracts.
Somebody on your team is running a platform evaluation right now: benchmark charts, feature grids, analyst quadrants. Here is the uncomfortable shortcut. The answer was mostly decided years ago, by three choices nobody thought of as AI decisions: which ERP you run, which CRM is your system of record, and what your private equity sponsor or strategic partner already standardized on. Your stack picked your archetype. The only question is whether you design with that gravity or pay to fight it.
The principle underneath this playbook is data gravity: the best model is the one your data already lives near. Agents are only as good as the context they can reach, and your context accumulated inside the systems you have been paying for since before the AI budget existed. Benchmarks compare models on neutral ground; your company is not neutral ground.
| Archetype | The tell | The strategy | The common error |
|---|---|---|---|
| The Microsoft shop | M365 is the daily operating system; the data layer is Azure. | Consolidate on Copilot, treat Fabric as the data moat, avoid agent sprawl. | Trying to build lean on the side. You pay for both and integrate neither. |
| The Salesforce shop | The CRM is the system of record; the customer moat lives in Data Cloud. | Build in Agentforce, use Data Cloud as the memory layer, resist point solutions. | Ignoring Data Cloud and bolting agents onto exports of your own data. |
| Lean and tech-forward | Minimal legacy, engineering culture, builds first. | Compose the constellation: prototype fast, run on the platform that fits each job. | Buying enterprise suite licenses you will never open, for optics. |
| Vertical-SaaS heavy | 60 to 80 percent of workflow runs inside one vertical system. | Pressure the vendor for AI roadmap commitments, and run one small open stack purely as negotiating leverage. | Waiting for the vendor to innovate. They ship on their schedule, not yours. |
The archetype names carry today's vendors, and the vendor names will age; the diagnostic does not. Whoever holds your systems of record holds your agent strategy, whatever they are called in three years. For vertical operators especially (the practice-management system, the construction platform, the policy system), the honest reading is the least comfortable: least control, most lock-in, and also the fastest time to value when the vendor ships something good. That is why the leverage stack matters: a small parallel open-source deployment exists not to replace the vertical vendor, but so that your renewal conversation has an alternative in the room.
The reason to name your archetype out loud is what it does to governance. One public example from the first edition still carries the whole lesson: one CEO walked into the board with a framework (archetype, first function, outcome metric) and put the vendor list in the appendix; when vendors changed over the following eighteen months, the framework held and the board approved each reallocation without drama. The counterexample spent tens of millions with the vendor's name as the strategy on slide two, and when the vendor failed there was no framework left to recover with. Frameworks compound; vendor lists go obsolete in eighteen months. Your archetype belongs on slide two. The logos belong in the appendix.
Naming the archetype costs pulling your top five AI-relevant contracts and one honest hour. Living by it costs discipline at purchase time: every non-integrating point solution someone wants is technical debt bought on purpose, and it should be defended as such or dropped. Fighting your archetype instead costs integration work forever, on every tool, for as long as you own both.
The vendor names in the archetype table have already been refreshed once since the first edition, and they will be refreshed again; that is what the as-of chip on this page is for. The archetype question itself has not moved at all, which is the point of teaching it this way. One emphasis did sharpen: the leverage stack for vertical operators went from a nice idea to the single most effective renewal tactic I have watched a mid-market company use.
The stack and data-gravity read is part of every Phase Zero assessment. Already know your archetype and your first build? A Pilot Build starts there.
Start with a Phase Zero assessment Explore all engagements