A hundred-person engineering org evaluated us last month. The shortlist wasn't us versus two other vendors. It was us, two other vendors, and paying an AWS partner to build the whole thing as an internal tool.
That fourth option is on every serious shortlist now, and most agent vendors are pretending it isn't. Coding agents made the first version of almost anything cheap. A competent platform team can wire up a working internal agent system in a few weeks, so at evaluation time the build option always looks credible. If your pitch only beats the other vendors, you lose to the invoice from the systems integrator.
What won the deal was not a feature. It was flexibility. We don't ship a proprietary CLI and demand the team abandon their existing tools — we wrap the harnesses they already use. That mattered more than anything on the comparison spreadsheet, because the buyer's real fear wasn't picking the wrong vendor. It was getting locked into someone's opinionated stack while the underlying models and tooling churn every quarter. Rigidity reads as risk. The platform that absorbs change on the customer's behalf beats both the rigid vendor and, critically, the internal build — because engineers keep building it themselves right up until they price in what happens after launch.
And what happens after launch is the actual argument. The internal tool works on day one. Then the champion gets reassigned, the model provider ships a breaking change, three teams ask for different defaults, and nobody owns the maintenance. Internal AI tools have a shelf life, and it's shorter than the procurement cycle that approved them. The vendor's job is to make that decay curve someone else's problem — yours — permanently.
If you're selling agent infrastructure, stop building the pitch deck against the other logos in your category. Build it against the buyer's own engineers. If you're buying, ask every vendor one question: what happens to this when the model landscape shifts again in six months? The answer tells you whether you're buying a platform or renting a snapshot.
Related Essays
Agent Build Versus Buy: Why Engineers Keep Building It Themselves
An engineer can ship a working agent harness in a week and a half. Code is cheap. So what does a paid agent platform offer that a week of engineering does not?
What the Build-vs-Buy Data Actually Shows
From Stripe to a five-person startup, the agent stack is mostly blue — built in-house. The harness is bought. The middle of the stack is built. The opportunity sits in turning blue dots green.
Homegrown Platforms Decay
Internal agent platforms are built by ambitious individuals with other jobs. When those engineers move on, the platform becomes a liability.
Key takeaways
- Every serious agent platform evaluation now includes the option of building the tool internally.
- Flexibility and harness-agnosticism won the deal — buyers fear lock-in to a proprietary stack while models churn quarterly.
- The vendor pitch has to beat the internal build on maintenance burden, not just on launch-day features.
FAQ
Why do buyers consider building agent tooling internally at all?
Coding agents have made the first version cheap. A competent team can stand up a working internal agent in weeks, so the build option always looks viable at evaluation time — the cost shows up later, in maintenance.
What actually differentiates a vendor against the internal-build option?
Flexibility and ongoing maintenance. A platform that wraps the buyer's existing harnesses and absorbs every model and tooling shift is hard to replicate in-house, where the internal tool decays the moment its champion gets reassigned.