Internal Mobility With AI: Match People Before They Look Outside

When a COO of a 220-person services firm told me she'd just lost a senior designer to a competitor — for a role she had open in marketing on the same floor — I asked if either side knew about the other. Neither did. The designer had spent two months talking to recruiters; the marketing lead had spent two months filtering external resumes.

Why don't most SMBs run internal mobility programs?

Because they don't have a People Ops team, and because the manual version is exhausting. Knowing every open role, knowing every employee's actual skills and interests, knowing which combinations are plausible — that's three full-time jobs of context spread across a founder, a head of people, and a handful of managers who don't talk to each other.

Definition: Internal mobility — the deliberate practice of moving existing employees into new roles inside the company instead of hiring externally, with a structured process that surfaces matches before people start job-searching.

Without a system, you only see the matches retroactively — after someone leaves for the role you already had. AI makes the matching cheap enough that it can run every quarter and surface candidates the founder didn't see.

What does an AI matching program actually do?

Three inputs, one output, one cadence.

Input 1 — Skills inventory

Self-reported skills, with manager validation. A short structured form, refreshed every six months. Not a 200-item taxonomy — a 20-30 skill list per role family, with a 1-5 self-rating and one example sentence per "high" rating.

Input 2 — Interest signal

What would you want to do in the next 12 months that's different from what you do today? One paragraph, voluntary, opt-in to the matching program. No interest signal means no inclusion — non-negotiable for the trust-not-surveillance principle.

Input 3 — Role openings

Every open req with the realistic skill requirements (not the wish-list version that filters everyone out). Posted internally before externally, with a 7-day internal window.

Output — Quarterly match report

Per opening, top 3-7 internal matches with: skill overlap %, interest alignment, manager-validation status, and a "conversation starter" line. The output goes to the hiring manager and the matched employee's manager simultaneously — never to the matched employee directly without their manager's awareness, and never as an automatic offer.

Cadence — Quarterly

Once a quarter, not continuously. Continuous matching is what makes the program feel surveillance-y. Quarterly is enough to catch matches before they leak; more often turns the workflow into a creepy "your manager got an alert about you" experience.

How do you avoid the surveillance trap?

Five rules.

Definition: Surveillance-style implementation — an internal mobility system that operates without employee opt-in, surfaces matches to leadership before employees know, or aggregates behavioral signals (Slack, calendar, code commits) to infer interests.

1. Opt-in only. No interest paragraph = no inclusion. The interest paragraph is the consent.

2. No behavioral mining. Do not feed Slack messages, calendar patterns, code commit history, or 1-on-1 notes into the matching system. The cost in trust is much higher than the marginal accuracy gain.

3. Match report goes to managers, not directly to employees. The matched employee's current manager sees the match first and decides whether to have the conversation. This protects employees from "your manager already knows you want to leave the team" awkwardness.

4. Right to decline silently. The matched employee can decline the conversation with no record kept beyond "declined." No "why did you decline" form. No follow-up matches forced into the next quarter.

5. No retaliation surface. A manager cannot use the match report as evidence in a performance conversation. Codified in the program description and enforced by HR.

If you can't hold these five rules, do not run the program. The damage from a surveillance-feeling program exceeds the value of the matches it surfaces.

Copy/paste prompt template — Quarterly internal match

You are matching internal employees to internal open roles for a 30-500-employee company.

Inputs:
- Open roles (this quarter): [LIST: title, team, required skills (must-have vs nice-to-have), location, level]
- Employee pool (opt-in only): [LIST: name, current role, current team, skills (1-5 self-rated + manager-validated tier), interest paragraph]

Output, per open role:
1. Top 3-7 matches, ranked.
2. For each match: skill overlap %, interest alignment (HIGH / MEDIUM / LOW with one-line evidence from interest paragraph), skill gaps (what would need development), tenure-in-current-role months.
3. "Conversation starter" line — one sentence the hiring manager could use if both sides agree to talk.
4. Caveats and risks per match.

Hard rules:
- Match only opt-in employees. Exclude anyone whose interest paragraph is blank.
- Do NOT use any behavioral data — only self-reported skills + interest paragraph + manager skill validation.
- If a match has tenure under 12 months in current role, flag for "is this team blockable?"
- If a match is the only senior on their current team, flag for "succession risk."
- Output goes to both managers; do NOT generate text for direct-to-employee use.

The "caveats and risks" line is what makes the report defensible. A match that surfaces a senior engineer who's the only senior on their team is a different conversation than a match that surfaces a junior with bench depth behind them.

Tool tip (AIAdvisoryBoard.me): Internal mobility data on its own doesn't tell you whether the program is working — you need to see how the management layer is responding to matches week-over-week. The Plan → Fact → Gap pattern shows which managers had the conversation, which let the match slip, and which teams are quietly building succession depth vs. losing their only senior. SMBs running this view post-match cycle see retention shifts within one quarter. See how the 7-day diagnostic works at https://aiadvisoryboard.me/?lang=en.

Manager scan (2-minute digest example)

• Plan: opt-in rate target 35% of employees; quarterly match cycle holds 90-day rhythm
• Fact: opt-in rate 41%; this quarter surfaced 18 matches across 6 open roles
• Gap: 4 managers did not act on matches within the 14-day window; 2 cited "wrong time"
• Conversations held: 11 of 18 matches led to a real conversation
• Internal hires: 4 from this quarter's matches; 3 still in discussion; 11 declined or paused
• Comp parity: all 4 internal hires received the same band offer they'd have gotten externally — no internal discount
• Time to fill: matched roles filled ~40% faster than non-matched comparable openings
• Surveillance complaints: zero opt-in concerns flagged in anonymous channel
• Manager training: 2 managers identified for "matching conversation" coaching after awkward debriefs
• Trust signal: opt-in rate up 6 percentage points cycle-over-cycle

Micro-case (what changes after 7-14 days)

A 160-person product company ran their first quarterly match cycle last summer. 42% opt-in rate, 12 matches across 5 open roles. Three internal hires inside 30 days — one senior PM moving to growth, one designer moving to product marketing, one engineer moving from infra to ML. The senior PM was actively interviewing externally and had a competing offer in week 2; the internal move closed the conversation. Externally, that backfill would have taken ~12 weeks and cost ~$30k in recruiter fees. Internally, the cost was a conversation, a 4-week ramp, and a backfill on a more junior role that was an easier external hire. Three quarters later, the program had filled 14 roles internally and the founder had stopped describing the company as "constantly bleeding senior talent."

Note on this case: This example is illustrative — based on typical patterns we observe with companies of 30-500 employees, not a single named client. Specific numbers are rounded approximations of common ranges, not guarantees.

Tool tip (AIAdvisoryBoard.me): The reason most internal mobility programs die isn't the matching — it's that managers quietly stop having the conversations after the second or third cycle, and nobody notices until the program is dead. Plan → Fact → Gap on manager conversation rate post-match-cycle is the single metric that keeps the program alive. The 7-day diagnostic gives you that view automatically, in the daily digest. See it work at https://aiadvisoryboard.me/?lang=en.

FAQ

What if a manager doesn't want to lose a matched employee? The conversation happens anyway. The losing manager gets a 30-day window to make the case for staying, including any growth-path commitment they're willing to put in writing. If the employee still wants to move, the move proceeds. Blocking moves silently is the fastest way to kill the program.

Should the matched employee see the report? No — they see it via their manager, who decides whether to surface the match. If you skip the manager and go direct, you're optimizing for short-term placement at the cost of long-term trust.

What about employees who never opt in? That's fine — opt-out is a feature, not a failure. Some people prefer to surface interest through direct conversation with their manager, and that's a valid path. Track opt-in rate as a trust signal; don't chase a target.

Can we use behavioral data to improve matches? You can. You shouldn't. The accuracy gain is small; the trust cost is large. Once employees suspect that calendar invites or Slack messages feed the matching system, opt-in collapses and the program is over.

What's the right cadence? Quarterly. Monthly feels like surveillance; semi-annual is too slow to catch attrition before external offers land. Quarterly hits the right balance for a 30-500-person company.

Conclusion

Losing a senior employee to a competitor for a role you already have open is the most preventable form of SMB attrition. An AI matching program — opt-in, quarterly, manager-mediated, behavior-data-free — surfaces matches before the resignation. The matching is the cheap part; the trust rules are what make the program survive past cycle three.

Pick your first quarter. Write the five trust rules before the prompt. Run the match. See who you almost lost.

If you want a system that surfaces the Plan → Fact → Gap on manager response to matches — automatically, every day — see how the 7-day diagnostic works at https://aiadvisoryboard.me/?lang=en.