Resources / How-to

How to use AI for your business

Six steps from "we should do something with AI" to a working system in production. Skip the parts you've already done; don't skip the order.

1. Pick one high-leverage workflow

Not three. One. The right pick is a workflow your team does dozens of times a week, where output quality varies, and where a 30% speed-up would be visible to the CFO. Examples: ticket triage, RFP responses, contract review, lead qualification.

2. Inventory the data and access

AI is only as good as what you can feed it. List the data sources, who owns them, and whether you can ground or fine-tune on them legally. If the answer is "we'd have to clean this for two months," pick a different workflow.

3. Choose build vs. buy vs. wrap

If a SaaS already does 80% of it, buy. If your moat is the workflow itself, build. If you need a custom assistant on top of your own data, wrap an API. Most "AI for business" projects are the third one.

4. Staff the right crew

You need an AI PM, an applied AI engineer, an MLOps engineer, and a domain SME. Pulling one engineer off another team and calling it "the AI team" is the #1 way these initiatives die. See typical crew shapes and costs.

5. Ship to real users in 8–12 weeks

Pilot to a small, friendly user group. Real usage exposes everything a demo hides: latency, hallucinations, edge cases, change management.

6. Measure, then scale or kill

Define the metric before you start. Time saved per ticket, conversion lift, error rate. If it moved, scale. If it didn't, kill it without shame and try the next workflow.

More context: the full AI for business guide and what AI can actually do for your business.