Consultant vs. internal AI hire: how mid-market operators should staff AI work

If you run a mid-market company and you’ve decided AI is no longer optional, the next question is uncomfortable: who actually does the work? Hire an internal AI lead? Engage an external partner? Both? The wrong call wastes a year and a quarter-million dollars. This guide is a decision framework for that choice. By the end you’ll know which path fits your situation, what to budget, and what to watch out for on either side.

It’s written for operators at companies doing $5M-$100M in revenue (the fit is strongest from $5M-$30M) — founder-led, owner-operated, family-owned, or PE-backed. If that’s you, read on. If you’re an enterprise with an established AI org, most of this won’t apply.

The core trade-off

There are three paths: hire, partner, or both. The honest trade-offs look like this.

Neither column is “the answer.” The answer depends on five questions.

The five decision questions

1. Have you ever shipped a successful AI project before?

If you haven’t, you don’t know what good looks like yet — which means you don’t know what to hire for. Bring on an internal AI lead before you understand your own needs and you’ll get a wrong-fit hire: either too senior (and expensive) or too junior (and slow).

• No → Partner first.
• Yes → Either path is viable.

2. How big is your AI workload over the next 12 months?

Internal hires need utilization. If you have two or three small initiatives in mind, you can’t justify a full-time senior salary against that workload.

• A few initiatives over 12 months → Partner.
• Multiple initiatives across multiple teams → Internal lead, possibly alongside a partner.

3. How fast do you need to start?

Hiring a senior AI lead takes four to six months. Then they need three to six months to ramp. That’s close to a year before they ship anything that matters. A partner starts in days and ships inside the first month.

• This quarter → Partner.
• Next year is fine → Either.

4. What’s your AI budget?

A credible internal AI team costs $400K+ per year. Below that number, you can’t fake it — you’ll hire one underpowered person and call it a program.

• Under $200K/year → Partner only.
• $300K-$500K/year → Internal lead solo, or an upgraded partner relationship.
• $500K+/year → Both is on the table.

5. Is AI a moat for you, or just parity?

If AI is core to how you’ll win — and you’re in an industry where AI capability genuinely differentiates — you’ll eventually want that capability in-house. If AI is mostly about keeping pace with peers, a partner-only arrangement is fine indefinitely.

• Parity → Partner is fine long-term.
• Moat → Plan to build internal capability eventually.

The three paths in detail

Path 1: Internal AI lead only

What it looks like: You hire a Head of AI, Director of AI, or AI Officer. They build a team over time and own both strategy and execution.

When it works:

• You have $400K+/year for a serious internal team.
• You have enough AI work to keep them fully utilized.
• You can wait six-plus months for ramp.
• AI is a competitive moat for you.

When it fails:

• You hire too senior (can’t justify the cost) or too junior (slow, error-prone).
• The hire leaves and you start over.
• You hire from outside your industry and they never ramp in time.
• The board changes priorities before they ship.

Real costs:

• Senior AI lead in Texas: $250K-$400K total comp
• Engineer to support them: $150K-$250K total comp
• Tools, training, conferences: $30K-$50K
• Total team cost: roughly $430K-$700K/year

Path 2: External partner only

What it looks like: You engage an AI Office retainer. The external team supplies senior judgment, engineering capacity, and ongoing operations. You supply an internal champion and an executive sponsor.

When it works:

• You’re early and don’t yet know what to hire for.
• Budget is under $200K/year.
• You want speed — starting in days, shipping in the first month.
• You want pattern matching from a team that’s done this across many companies.
• AI is parity, not moat.

When it fails:

• You can’t name a champion or sponsor internally.
• You don’t trust an outside team with your data and processes.
• You pick a partner who sells projects, not an ongoing relationship.
• You pick a partner who’s too junior or too generalist.

Real costs (the offering ladder):

• AI Office — Sherpa: $2,500/mo (~$30K/year)
• AI Office — Operator: $5,000/mo (~$60K/year)
• AI Office — Embedded: $10,000/mo (~$120K/year)
• Value Sprints as you scale: $2K-$25K each, up to roughly $95K for the larger ones, each carrying a 12-month KPI guarantee
• Plus tooling (varies)

A note on how the economics work: the AI Office retainer is built to target at least 3X payback — it’s designed to pay for itself, and we structure it that way. We share risk on both. On Value Sprints and multi-quarter programs, our fee is tied to the agreed KPI — miss it inside 12 months and we keep working at our cost until it’s hit. On the AI Office retainer, we commit to the KPIs we set with you and keep working until they move (and you’re month-to-month, so we earn the renewal every month). Different structures, same principle: we own the outcome.

Path 3: Both

What it looks like: You have an internal AI lead and an external partner. The internal lead owns the daily operating cadence; the partner provides senior strategic input, surge engineering capacity, and broader pattern matching.

When it works:

• You’re past the early stage and AI is operational.
• You have multiple initiatives running in parallel.
• Your internal lead is still ramping and the partner accelerates them.
• You’re PE-backed with AI needs across multiple platforms.
• You want a hedge against single-point-of-failure on AI strategy.

When it fails:

• You never defined who owns what, so partner and internal lead collide.
• The internal lead resists outside help.
• The partner gets defensive about scope.
• You’re paying for both and fully using neither.

Real costs:

• Internal lead and team: $400K+
• AI Office partner at a lower tier: $30K-$60K
• Total: roughly $430K-$460K+

What most mid-market companies should actually do

For most companies in the $5M-$100M range (strongest fit $5M-$30M, with $1M-$10M EBITDA), the sequence that works looks like this.

Year 1 — Partner only (AI Office). $30K-$120K/year. Senior team. Ship in the first 30 days. Build internal awareness. Document ROI carefully.

End of year 1 — Decide with data instead of guesswork.

• AI is delivering 3X+ payback and you have more in the pipeline → consider hiring an internal lead in year 2.
• AI is delivering payback and demand is stable → stay partner-only.
• AI isn’t delivering → change partners or pause. (A good partner will tell you when the honest answer is “not yet.”)

Year 2 — Add internal capability if it’s justified. Hire an internal lead if you’ve concluded AI is a moat. Keep the partner at a lower tier for senior judgment and surge capacity.

Year 3+ — Internal lead plus partner is the steady state for most successful mid-market AI programs.

What to watch out for

Red flags in vendors

• “You need a full-time AI team” — said without diagnosing your situation. That’s an upsell.
• “You don’t need internal capability, ever.” They’re protecting their billings.
• “We’ll transfer knowledge to your team” — with no concrete plan attached. They won’t.
• Hourly billing with no fixed-price option. They’re motivated to drag the work out.
• No published pricing. They charge whatever the market will bear.

Red flags in internal candidates

• Strong AI credentials, no business-operations experience. They’ll ship clever things that don’t move the number.
• Strong operations background, no AI delivery experience. They’ll over-promise and under-ship.
• Came from a much larger company. Mid-market scrappiness will surprise them.
• Wants to “build a team” before they’ve shipped anything. Wrong order.

Ten questions to ask any AI partner

1. How many mid-market clients have you delivered AI work to in the last 18 months?
2. Show me a working solution you shipped — not a deck.
3. Who is the actual delivery team? Names, backgrounds, where they’re based.
4. What does a typical first 90 days look like?
5. How do you measure ROI?
6. What happens when AI gets something wrong in production?
7. What’s your cancellation clause?
8. How do you transfer knowledge to my team?
9. What’s the longest-running client relationship you have, and why?
10. What kind of work would you turn down?

Ten questions to ask any AI hire candidate

1. Walk me through an AI project you shipped that produced documented ROI.
2. What did you get wrong on it, and what would you do differently?
3. How would you build out our AI capability in the first 90 days?
4. How would you decide what not to do?
5. How do you work with operations, finance, and sales when they don’t speak AI?
6. What’s the smallest thing you’d ship in your first 30 days here?
7. How do you handle it when AI gets something wrong in production?
8. What’s your working relationship with the C-suite, IT, finance, and ops?
9. What kind of AI work do you find boring but necessary?
10. What’s the next AI thing you’d hire someone else to do because you’re not the right person for it?

The honest bottom line

For most mid-market companies, the right answer for year 1 is a partner, not a hire. It’s faster, cheaper, lower-risk, and the pattern-matching advantage is real. After year 1, the decision gets made on data instead of guesswork — and that’s worth a lot.

If you’re partner-shopping, ask the ten partner questions. If you’re hiring-shopping, ask the ten candidate questions. If you’d like to walk through your specific situation — including the case where the right answer is “don’t do AI yet” — start a conversation with us.