Most learning organizations can’t answer a deceptively simple question: Who actually decided to automate the things you’ve already automated?
Not who approved the budget. Not who signed the vendor contract. Who made the call that this piece of how your people learn and grow should be handled by a machine instead of a human?
In most organizations, the honest answer is that no one did — at least not deliberately. And that should concern us more than it does.
Automation by Accretion
We like to believe automation is a strategic choice. That someone weighed options, considered implications, and made a thoughtful decision about what AI should own and what should stay human.
That’s a comforting story. It’s rarely true.
What actually happens is quieter. Someone finds a tool that saves them a few hours and starts using it. A vendor ships a new “AI-powered” feature, and suddenly it’s part of the platform you already pay for. IT selects a system based on technical specs, and the automation choices come preloaded before the people who understand the work ever weigh in. A budget gets approved for efficiency, and efficiency quietly becomes the criterion for everything.
None of these moments feel like decisions. They’re accumulations — small, invisible choices that add up to a strategy no one ever consciously chose.
A client we worked with recently illustrated this perfectly. They asked us to assess their AI readiness, assuming the answer would be about what to buy next. Six weeks later, the pattern was unmistakable: they already owned most of what they needed. Tools were purchased, half-activated, and quietly abandoned. In one case, multiple departments were requesting a system another department already had — and wasn’t using.
No one had decided not to use it. No one had decided anything. The tools arrived, got partially turned on, and the deciding simply never happened.
Automation by accretion isn’t an edge case. It’s the default state of most organizations. The automation is happening. The deciding is not. And when no one owns the decision, no one is accountable for its consequences.
That’s the part that matters. Not that things are getting automated — that’s inevitable. But that they’re getting automated, or half-automated and abandoned, without anyone answerable for whether it’s the right thing to automate, for the right reasons, with the right people in mind.
Why Learning Is Different
You could argue that automation-by-default is fine. Markets sort it out. People adopt what works and ignore what doesn’t. Why impose deliberate decision-making on something that seems to self-organize?
Because learning isn’t neutral territory.
When you automate a piece of how people learn, you’re not just making an operational choice. You’re making a decision about who gets developed, how, and how well.
Inside that same client organization, the learning function was a microcosm of the problem. Their platform already supported the skills structures leaders were asking for — but those features sat unactivated. Their skills data lived on a single spreadsheet. Their enablement plan existed on paper while the capability underneath it hadn’t been built.
Every one of those was a decision nobody consciously made: to leave capability dormant, to let critical data depend on one file and one person, to treat a written plan as though it were a built one.
And the most telling example wasn’t about learning content at all. They had launched a digital self-service capability — the kind of automation everyone applauds. But frontline staff weren’t credited when a customer they helped later moved to the automated channel. The very people whose cooperation the automation depended on had a quiet reason not to cooperate. No one had decided to disadvantage them. The incentive question simply never had an owner.
That’s what an unowned automation decision looks like. Not a dramatic wrong call. A capability that quietly works against the people expected to adopt it because the human consequences were nobody’s job to think about.
Now scale that across a learning organization. When one population gets an AI chatbot for development questions and another gets human coaching, you’ve made a decision about whose growth warrants human attention. When one function’s content gets automated because it’s high-volume and another’s stays bespoke, you’ve decided where to invest human craft. When a system recommends who’s “ready” for advancement based on historical patterns, you’ve handed a capability decision to a model trained on a past that may not deserve repeating.
These aren’t technical choices. They’re choices about workforce capability and, whether we name it or not, about equity — who gets access to the good stuff, whose development gets optimized away, whose context the automated system serves well, and whose it quietly fails.
Learning leaders understand this in a way budget holders and platform vendors often don’t. Which is exactly why they need to be in the room when these decisions get made — and why it’s a real problem that they often aren’t.
The Governance Gap
Most organizations serious about AI have built governance around it — privacy, security, compliance, risk, appropriate use. The machinery exists.
But look closely at what that governance covers, and you’ll see a gap. It’s almost entirely about protecting the organization from harm. Very little of it is about the effect of automation on the people inside the organization — their capability, their access, their development.
We govern AI for risk. We don’t govern it for human growth. And in learning, human growth is the entire point.
That gap exists because the people who built the governance frameworks were thinking about the organization as an entity to be protected, not as a collection of people whose development is being shaped by these choices. It’s not malicious. It’s a blind spot — and one learning leaders are uniquely equipped to fill.
What Decision Ownership Actually Requires
So, what do you do about it? Not build another committee. Please, not another committee. Diffusing a decision across a governance board is often just a sophisticated way of ensuring no one owns it.
Real ownership requires three things — simple, but uncomfortable.
1. Name who owns which decisions
Not “the AI steering committee.” A person. When an automation choice affects workforce capability or access, someone specific should be accountable for that call — able to explain the reasoning and answer for the outcome.
Ownership that disappears when one person leaves isn’t ownership. It’s a single point of failure wearing ownership’s clothes.
2. Separate consequential decisions from operational ones
Most automation choices don’t need ceremony. Automating meeting summaries or first-draft formatting is operational — let people move fast.
But automating who gets development, how capability gets assessed, or which populations receive human attention — those are consequential. They shape people’s futures. They deserve explicit ownership.
And there’s a distinction worth borrowing from the best implementations: every automated capability needs an adoption owner, not just an implementation owner. Implementation is who turned it on. Adoption is who’s accountable for whether it actually gets used, by whom, and to whose benefit.
3. Make the equity questions concrete
“Equity” as a slogan is useless. What matters is whether someone asked specific, checkable questions:
Who benefits from this automation and who bears the cost? Whose experience did we design for, and whose did we assume? Does the data behind this recommendation represent the people it’s being applied to?
Equity isn’t a value you declare. It’s a set of questions you either ask or skip.
The Tension I’ll Name Honestly
There’s a real risk here. If every automation choice requires review, you can build a system so cautious it can’t move — losing the responsiveness that made automation worth pursuing.
I’m not arguing for centralizing every decision or slowing everything down. Most automation should stay distributed and fast.
The argument is narrower: The small subset of automation decisions that materially shape workforce capability and equity should not be made by default. Those decisions deserve a name attached to them and a few real questions asked before the fact.
And one more honesty: treating equity as a box to check is worse than ignoring it. If we invoke it, it has to show up as specific questions with real answers — or we shouldn’t invoke it at all.
Why This Falls to Learning Leaders
Someone is going to make these decisions. The only question is whether they’re made deliberately by people who understand what’s at stake, or by default by people optimizing for cost and speed because no one gave them a reason to optimize for anything else.
Learning leaders are the natural owners — not because of title, but because of perspective. You understand that developing a workforce is different from processing one. That capability compounds or erodes based on choices that look small in the moment. That a decision framed as “efficiency” can quietly become a decision about who the organization believes is worth investing in.
If you’re not at the table when automation decisions get made, those decisions still get made — just by people asking narrower questions than you would.
So the work of this moment isn’t technical. It’s claiming ownership of decisions that are currently happening to your organization rather than being made by it. Naming who decides. Separating what’s operational from what’s consequential. Asking the equity questions out loud, before the fact, while they can still change the outcome.
Automation by accretion is the path of least resistance. It’s also the path where no one is responsible for what your organization becomes. You can do better than default. The question is whether you’ll decide to — or let the decision get made for you.
