Picking the wrong AI consultant is one of the most expensive mistakes a mid-market company can make right now. Most are either ex-Big-4 generalists who add "AI" to their decks or ex-engineers who want to build things you do not need. This is the buyer's guide we'd want a CEO to read first.
The fastest filter: ask for two clients you can talk to where they implemented AI, not just advised on AI. "Implemented" means the client has tools in production, used daily, with measurable outcomes.
If they cannot produce two such references, they are an advisory shop, not an implementation shop. Both have a role — but you almost never want pure advisory at this stage. You need someone who has gotten their hands dirty.
If the firm pitches you a 6-week "AI strategy" engagement that produces a roadmap deck and no working software, walk. The deck is not the deliverable. The deliverable is something working in your business.
AI is the tool. Your business is the work. A consultant who knows AI deeply but does not know B2B sales motions, content production, or customer success workflows will spend 3 months learning your business while billing you for it.
Ask them to describe — without notes — how a B2B SaaS sales cycle works at a 30-person company. If they cannot, they will not know which workflows AI should touch.
Bad consultants build whatever you ask for. Good consultants tell you which 30% of what you asked for is the right scope, and which 70% should be dropped or sequenced later.
In your evaluation conversation, ask: "Tell me about a recent engagement where you advised the client NOT to do something they wanted to do." If they cannot give a specific example, they will not say no to you either.
| Model | Typical price | When it fits |
|---|---|---|
| Fixed-fee audit/diagnostic | \$1,500-\$15K | Start here. Produces a written roadmap. Limits risk while you evaluate the firm. |
| Fixed-fee implementation | \$5K-\$50K | Best for shipping the first 1-3 workflows. Clear deliverable, clear scope. |
| Monthly retainer | \$3K-\$20K/month | For ongoing partnership after initial implementation. Should have defined deliverables. |
| Project-based with engineering | \$30K-\$300K | Only for custom builds. Most teams should not start here. |
| Hourly | \$200-\$600/hr | Avoid for AI strategy work. Misaligns incentives. Use for tightly-scoped one-off questions only. |