A practical playbook, built from actually running this exercise with clients, for spotting flight-risk roles, mapping your bench, and drafting development plans with AI so succession planning stops being the spreadsheet nobody opens.
I ask a version of this in almost every workshop I run: raise your hand if your company has a real succession plan. Out of thirty people, maybe two hands go up. The data backs up what that feels like from the front of the room: only 21% of HR professionals say their organization has a formal succession plan, and 56% have none at all [1]. Only 20% of HR leaders say they have someone ready right now to step into their most critical roles [2]. AI won’t fix a culture that avoids this conversation. It will turn a messy list of names and job titles into a flight-risk map, a talent inventory, and a first-draft development plan before your coffee gets cold.
Nobody wakes up excited to build a succession plan. Honestly, in workshops I put it right up there with cleaning out the shared drive: important in theory, permanently at the bottom of the list in practice. I’ve sat through enough leadership offsites now to know the pattern by heart. Someone raises succession planning in Q1. Everyone nods, it gets a line item on a slide somewhere. By Q3 it’s quietly gone, because payroll broke, or a product launch ate the calendar, or a reorg nobody saw coming swallowed the quarter whole.
The numbers back up what that feels like from the front of the room, and they’re worse than most people guess when I put them on the spot. Only 21% of HR professionals report having a formal succession plan in place. 56% have no plan at all, mostly because nobody has the time or the resources to build one [1]. That’s most companies flying blind on the one question that actually matters: who runs things when a key leader quits, retires, or gets poached.
It gets worse the higher up the org chart you look. DDI’s 2025 Global Leadership Forecast found that only 20% of HR leaders say they have leaders ready right now to fill their most critical roles, even though 75% of organizations say they’d rather promote from within than hire externally. On average, internal candidates can only fill 49% of critical positions immediately if a vacancy opened today [2]. Companies want to promote from within. They mostly can’t, because nobody built the pipeline that would let them.
My honest read, after running this exercise with a lot of teams: succession planning doesn’t fail because HR doesn’t care. It fails because the traditional process is slow and manual and political enough that it rarely survives contact with a busy quarter. AI doesn’t touch the politics, I want to be upfront about that. What it fixes is the slow, manual part, and that’s most of the reason plans die before anyone finishes them.
Let me set expectations before you open a new ChatGPT tab, because a good chunk of every workshop I teach goes to exactly this moment. The question I get most, almost every single session, is some version of just tell it who should get the promotion. That’s the wrong question, and I say so bluntly in the room. AI isn’t going to tell you that Priya is ready for the Director of Ops role and Marcus isn’t. It doesn’t know Priya, it doesn’t know your org’s politics, and it has no idea Marcus quietly resents being managed by his former peer. Those are human judgment calls. They should stay human.
What AI is genuinely good at is the unglamorous middle of the process: organizing scattered information, spotting patterns across a messy list of roles and names, drafting a first version of a plan you can then argue with, and asking you the questions you forgot to ask yourself. Think of it as a very fast, very literal analyst who never gets tired of reformatting a spreadsheet, which, if you’ve ever done this by hand, is worth more than it sounds.
Here’s a distinction that trips up almost everyone the first time: people assume feeding an AI more data produces a better recommendation, the same way more data helps a forecasting model. For a succession call, it doesn’t work that way. More data just gets you a more confidently wrong ranking, because the AI still can’t see the politics, the rough personal year someone just had, or the quiet resentment between two people who used to be peers. That’s the caveat I repeat most in workshops, and people usually only really get it after they’ve seen it happen once.
If you’re new to using AI for people decisions generally, it’s worth reading up on how to be a better manager with AI before you touch anything succession-related. The guardrails are the same: AI drafts, you decide, and nothing goes into a system that would embarrass you if an employee saw it.
Before you think about successors, figure out which roles actually need one. Not every job on the org chart is a succession risk. A flight-risk role is one where, if the person left tomorrow, the business would genuinely struggle: a single point of failure, a role with specialized knowledge, or a position that’s historically hard to backfill externally.
Open ChatGPT or Claude and paste in your org chart, or just a simple list of roles, tenure, and how replaceable each one is in your judgment. Ask it to help you sort, not decide. A prompt like this works well:
“Here’s a list of roles on my team with tenure and a rough note on how specialized each one is. Help me build a flight-risk matrix scoring each role on business impact (1-5) and difficulty to replace externally (1-5). Flag anything that scores high on both. Ask me clarifying questions if you need more context before scoring.”
I’ve watched a couple hundred people run some version of that exact prompt in workshops, and the follow-up questions the AI asks almost always catch people off guard: whether a role depends on a single vendor relationship, or a piece of undocumented process knowledge nobody wrote down. That back-and-forth is the real value, not the matrix itself. You walk away with a short list of roles worth investing real time in, instead of trying to build a succession plan for all 40 people on the team at once, which is how these projects die.
Once you know which roles matter, you need to know who could plausibly grow into them. This is where most spreadsheet-based succession plans quietly rot, because nobody updates the tab after the first workshop. AI won’t fix people’s laziness about updating a spreadsheet, but it will make building the first version dramatically faster.
Pull together whatever you already have: performance review notes, project history, skills self-assessments, manager comments from 1-on-1s. Feed the relevant, appropriately anonymized details into Claude or ChatGPT and ask it to build a structured talent inventory: name, current role, key strengths, stretch experience, and any stated career interest. Ask it to flag where you’re missing information rather than guessing.
I watched a mid-size logistics company do this in a single afternoon. Their HR lead had eighteen months of scattered performance notes sitting in three different systems. She exported what she could, ran it through an AI tool with a clear prompt about what fields she needed, and had a working talent inventory draft before lunch. It wasn’t perfect. But it was ten times further than the blank spreadsheet she’d been staring at for a year.
I’ve seen the same thing play out at a regional bank, minus the happy ending, worth mentioning here as a caution. A branch operations lead tried the same approach but skipped the anonymizing step and pasted full performance reviews, names, salaries, the works, straight into a general consumer AI tool. Nothing catastrophic happened, but her compliance team was not thrilled when they found out, and she had to redo the whole exercise through an approved internal tool. Fifteen extra minutes of stripping out sensitive fields upfront would have saved her two weeks.
If your team already runs structured employee engagement surveys, that data is a goldmine here too. Career-interest questions and engagement signals feed directly into who’s worth developing next.
Having a name next to a role isn’t the same as that person being ready. This is the step most manual succession plans skip entirely, because it requires comparing what a role actually needs against what a candidate currently has, role by role, which is tedious by hand and takes minutes with AI.
Give the AI two things: a rough competency profile for the target role (the skills, experience, and scope it actually requires) and what you know about the candidate today. Ask it to identify the gap in plain language, not a vague “needs more experience” but specifics: has never managed a budget over $500K, hasn’t led a cross-functional project, has no exposure to the board.
Let’s be honest: most “high potential” lists are really just people the current leader likes and talks to a lot. Running an actual gap analysis, even a rough AI-assisted one, forces you to name the specific thing missing instead of vibes. That alone catches people who were overlooked and thins out candidates who were overrated.
This is also where DDI’s research is worth remembering: leaders consistently overestimate readiness when they rely on gut feel and current performance instead of a structured look at future-role requirements [2]. A five-minute AI-assisted gap check is a cheap way to add a bit of that structure without buying an enterprise assessment tool.
A gap you’ve identified but done nothing about isn’t a succession plan. It’s a worry list. Once you know what someone’s missing, AI is genuinely useful for drafting the actual development plan: stretch assignments, specific training, mentoring pairings, and a realistic timeline.
Ask ChatGPT or Claude to draft a 6-12 month development plan based on the gap you identified, and be specific about constraints: budget, whether the person can realistically take on a stretch project right now, whether there’s a mentor available internally. Then edit it hard. I’ll admit I once let an AI-drafted plan go into a 1-on-1 without editing it enough, early on. It read like a robot wrote it, because one had, and both the manager and the employee could tell. Lesson learned the annoying way. The AI draft is a starting point, not a document to hand someone unchanged.
Picture a manager who’s identified that her strongest analyst could grow into a team lead role but has never managed conflict or run a hiring process. An AI-drafted plan might suggest: co-lead the next hiring round with her manager, take a short course on difficult conversations, shadow a cross-functional project lead for a quarter. None of that is revolutionary. It’s organized, specific, and written down instead of vaguely promised in a hallway conversation, which is usually where these plans actually go to die.
This pairs well with how you already run 1-on-1s and onboarding conversations, since a development plan only works if it’s revisited regularly, not filed away after one meeting.
If you’re a small team, ChatGPT or Claude plus a shared doc is genuinely enough. But if you’re managing succession across dozens or hundreds of roles, dedicated HR platforms have built real AI features into this exact workflow, and it’s worth knowing what they actually do before you evaluate a purchase.
None of these tools make the judgment call for you. What they replace is the spreadsheet, the one that always lived in exactly one person’s downloads folder and nobody else could find. If your organization already runs one of these platforms, you probably have more succession-planning horsepower sitting unused than you realize. Worth asking your HRIS admin what’s actually turned on before you go buy something new.
A few patterns show up over and over when leaders start using AI for this kind of work, and they’re worth naming plainly.
Succession planning was never really a technology problem, if I’m honest. It’s a discipline problem: the plan dies when nobody protects the time to keep it current. AI removes the excuse that the first draft takes too long. It doesn’t remove the need to actually sit down every quarter and look at it again, which is the part everyone still skips.
It can help you organize the raw material of one: flight-risk roles, a talent inventory, readiness gaps, draft development plans, much faster than doing it by hand. It can’t make the final call on who’s ready for a role, since that requires context about people and politics that only you have. Use it to get a draft moving, then make the call yourself.
Not without checking your company’s data policy first, honestly, I wouldn’t risk it otherwise. Avoid pasting sensitive details like compensation, medical information, or full performance review text into a general consumer AI tool. Strip inputs down to what’s necessary, role, tenure, general strengths, skill gaps, or use an enterprise AI tool with the data protections your company has already vetted.
Replacement planning is the reactive version, basically a fire drill: who covers this role if someone quits tomorrow. Succession planning is slower and proactive, developing someone now so they’re genuinely ready when a role opens, not just scrambling to cover it. AI doesn’t change that distinction. It does make the proactive version a lot less painful, because the readiness-gap analysis and development-plan drafting that used to eat an afternoon now takes minutes.
For a small team or a handful of critical roles, ChatGPT or Claude plus a shared document is genuinely enough to get a real succession plan off the ground. Dedicated HR platforms like Workday, SAP SuccessFactors, Lattice, and Eightfold become worth the investment once you’re managing succession across many roles and need the data connected automatically to performance and compensation systems.
Quarterly is a reasonable default for most teams, with an ad hoc review any time a critical role’s incumbent changes, a reorg happens, or a high-potential employee’s circumstances shift. Because AI cuts the time it takes to refresh a flight-risk map or readiness gap, there’s less excuse to let a plan sit untouched for a year, which is still the single most common way succession plans go stale.
I put this together the way I put together most of what I teach in workshops: I started from DDI’s Global Leadership Forecast 2025 and its succession planning best-practices writeup and SHRM’s succession planning toolkit, then went and actually opened the current AI succession-planning product pages from SAP SuccessFactors, Eightfold, and Lattice to see what those features look like today rather than trust a marketing summary from a year ago. I cross-checked all of it against my own notes from running succession-planning sessions with corporate teams, including a few of the specific missteps I’ve watched people make live in the room.
Weekly tips. Real examples. Practical help for busy professionals.
We care about your data, check out our privacy policy.