Call prep — BIL's new role at Banyan

His situation: stepping in to lead the virtual care division, overseeing 1–2 software engineers / technical directors. Wants advice on (a) assessing talent, capabilities, and gaps, and (b) AI tools and capabilities broadly. He is (presumably) not an engineer — that's fine, and the advice below assumes it.

1. Assessing technical talent when you're not technical

Judge by artifacts and explanations, not tech trivia. With a 1–2 person team there's no room for proxy metrics — go straight to the work:

• "Walk me through the last thing you shipped — from idea to production." The quality of the explanation to a non-engineer IS the assessment. Strong engineers make complexity legible; weak ones hide behind jargon or blame the stack.
• Look at the exhaust, not the claims: how often does the team ship? What happened during the last incident/outage? Is anything written down (docs, runbooks), or does it all live in one person's head?
• Bus factor is the #1 risk at this team size. Ask each person: "what breaks if you're out for two weeks?" If the answer is "a lot," that's the first gap — before any AI conversation.

First 30 days = listening tour with written outputs. Ask each technical lead for a one-page "state of the system": what's solid, what's fragile, what keeps them up at night. Two payoffs: an instant map, and a calibration instrument — compare what each person flagged against what actually breaks over the next quarter. That tells him whose judgment to weight.

2. Capabilities & gaps — map against the mission, not a skills matrix

Skip the generic skills-matrix exercise. Instead: what must the division ship in the next 12 months? Work backward from that list to capability gaps. The filter for whether a "gap" is real: does it block a committed deliverable? If not, it's noise. (This also gives him a defensible story for hiring asks — "we can't deliver X without Y" beats "we should have Y.")

3. AI in virtual care — constraints first, then leverage

• The constraint story leads: PHI/HIPAA. No patient data in consumer AI tools, period. Anything touching PHI needs a BAA-covered platform (the enterprise tiers of the major AI vendors and cloud providers offer BAAs now — but it has to be the covered deployment, not an employee's personal account). His first policy act should be a one-page "what's allowed where" so the team isn't guessing.
• Highest-ROI safe lanes in healthcare right now: ambient clinical documentation/scribing, admin/prior-auth drafting, internal knowledge retrieval over policies and protocols. These are proven categories, not science projects.
• His biggest immediate lever is probably engineering acceleration. A 1–2 person eng team with modern AI coding tools (Claude Code and peers) is a genuine force multiplier — this is the cheapest capacity he can add, and it's PHI-free.
• Sequencing advice: pick one narrow, measurable use case with a real bottleneck behind it and land it before going broad. Adoption theater (60 pilots, zero production) is the standard failure mode.

4. Framing gift Ben can offer

Ben does capability assessment professionally — the useful gift for the call isn't tool names, it's the method: inventory what the division must deliver → rank candidate AI use cases by value × feasibility → check each against the constraint (PHI) → ship the top one and instrument it. Offer to be his sounding board when he gets to the ranked list.

Open facts (don't assert on the call without checking)

• Which "Banyan" — not verified (Banyan Treatment Centers / Banyan Health Systems / Banyan Medical Systems are all healthcare orgs with virtual-care surfaces). Location and org context change the advice at the margins; Ben will learn this on the call.
• BIL's technical depth — unknown; the artifact-based assessment approach works either way.