9 minute read
We have an obvious interest in you hiring us. So take what follows as our honest attempt to tell you what you should actually be evaluating — including things you should use to evaluate us.
The AI consultancy market is noisy. A lot of firms are selling AI strategy that amounts to workshops and slide decks. A smaller number are building things that work. Here is how to tell the difference.
The Checklist
Can they show you deployed systems, not demos? Demos are easy to build. Production deployments are hard. Ask for references from clients whose systems are live, handling real volume, and have been running long enough to encounter edge cases. A consultancy that cannot provide this is selling potential, not capability.
Do they talk about governance before they talk about features? Any experienced practitioner of agentic AI will lead with questions about scope, access control, escalation paths, and auditability. If a consultancy's first conversation is entirely about what the agents will do and not at all about how you will oversee them, that is a warning sign.
Do they have a view on where AI should not be used? Trustworthy consultants will tell you when a workflow is not a good candidate for automation. If every problem you describe gets a "yes, AI can do that" response with no pushback, you are talking to a vendor, not an advisor.
Can they measure the outcome? ROI is not a secondary concern — it is the primary one. Ask how the engagement will be evaluated. If the answer is vague, the engagement will be too.
Do they understand your industry's specific constraints? Data handling, compliance, integration with existing systems — these are not generic problems. A consultancy that has not asked about your specific regulatory environment or technical stack before proposing a solution is proposing the wrong solution.
Are they honest about timelines? Agentic systems take time to build correctly. A firm that tells you a production-ready deployment will be done in two weeks is either planning to cut corners or has not scoped the work seriously.
Why We Are Telling You This
Because we pass this checklist. We can show you NamingForce, EchoTexting, and Grademate. We talk about governance first. We tell clients when automation is the wrong answer. And we measure every engagement against verifiable outcomes.
If you want to put us through this checklist directly, we welcome the conversation.