Tech isn’t a business goal

Seen on LinkedIn:

Client: I need AI

Junior AI Engineer: Cool! Let's build a multi-agent system.

Senior AI Engineer: What's your business goal?

Client: Query our SQL database in plain English.

Junior: Okay, multi-agent is the way to go!

Senior: One LLM + Schema will do.

The point: don't over-engineer. Ask about business goals first. So far so good.

But "query our SQL database in plain English" isn't a business goal. It's a technical capability that may or may not enable one.

The client has self-diagnosed (fine—it's their business) and then self-prescribed (dangerous when the remedy is outside their expertise). Our job is to ask deeper questions. What's the real problem behind the presenting problem? What's the business goal behind the ask for a technical capability?

Hypothetical continuation:

Client: Query our SQL database in plain English.

Senior: What would that unlock in your business?

Client: Our sales team wouldn't waste three hours each day waiting on data analysts to pull reports!

Senior: What sort of reports?

Client: Yesterday's sales data for each store, grouped by item category.

Senior: That's a simple dashboard. No AI required. We'll have it ready by tomorrow.

But what if instead:

Senior: What would that unlock in your business?

Client: Regional managers need to cross-reference sales, inventory, supplier lead times, and support tickets to make judgment calls. Every situation is different—last week someone had to figure out which stores could absorb inventory from a closing location.

Senior: Unpredictable queries across multiple domains. Multi-agent might actually be the right call here.

Simple isn't always best. The only way to know is to keep asking questions—especially business questions.