AI has made analysis fast. It has not made judgement fast, and it should not. What it has done is pile far more analysis at the door of the few people expected to judge it.
McKinsey has now named this well. In a new McKinsey Quarterly piece, Alex D'Amico and colleagues describe a strategy team that used AI to turn five weeks of financial analysis into less than two days. The decision did not arrive any sooner. The leaders still needed time to understand the assumptions and the data behind the numbers. The authors put it plainly: analysis now runs on AI time, and judgement still runs on human time (McKinsey, 2026).
They are right that this gap is the new constraint. They are right, too, about what cannot be compressed: the human judgement and the alignment building that turn analysis into a sound decision, to avoid the risks that arise when the two operate at different speeds. The question worth asking is what to do with that human time, once you accept it cannot be shortened.
The usual fix keeps judgement in a few heads
McKinsey's recommendations are sensible. Keep senior attention for the rare, high-stakes calls. Let AI support or run the routine ones inside clear limits. Train people to check what the AI produces. Make one person accountable for each critical decision.
Look at what these have in common. Judgement is treated as something that happens in a small number of heads near the top. The fix is to protect those heads, ration their attention, and train the people who feed them. For routine calls that works well. For the rare ones it leaves the bottleneck exactly where it was, with a growing queue in front of it.
The authors see the risk this creates. As the volume of output grows, AI ends up being used to assess its own work, with less human judgement and less testing of risk along the way. A queue that cannot be cleared by people gets cleared by the machine.
Accountability should stay with a person. Judgement does not have to.
The rare decisions depend on what no data set holds
By McKinsey's own split, the decisions that need human judgement are the rare, high-stakes ones: entering a new market, repricing in a disruption, planning for a future when past patterns no longer hold. What makes them hard is that historical data cannot fully capture what they depend on. For consequential strategy decisions, people bring a different set of strengths to the process: interpreting evidence and testing assumptions, recognising what wasn't anticipated, building alignment around decisions, and mobilising the organisation to execute. So allow AI to cultivate less judgement.
So where does the judgement for them come from? Not from a better model. Not from a senior leader reading faster. People don't build judgement through passive consumption of finished answers. It forms when people who see the situation from different places think it through together: the sales lead who hears customers hesitate, the engineer who knows what the platform can really carry, the partner who sees the market from outside, the team that knows what the regulator is about to ask.
This is not a matter of collecting what each of them already knows. The understanding is built in the exchange. One person's point changes what the next one sees, a doubt meets a piece of evidence, and an implication nobody had named appears. That is the judgement the rare decisions need, and it does not sit in any one head waiting to be rationed.
Give human time to more of the right humans
If judgement runs on human time, the lever is not less time. It is more of the right people, each contributing in their own time. A leadership team working through AI output alone spends its hours in a queue. The same kind of hours, spread across the people who carry the decision, become a deliberation.
That is what collective sensemaking does. On Hunome, contributors add Sparks to a SparkMap that stays open for as long as the decision matters: a point, a doubt, a piece of evidence, an implication. Each Spark is characterised by the person who makes it: the knowing behind it, its stance, what it says is missing, the future it points to. Others build on it, connect to it and challenge it. Nobody has to be in the same room at the same hour, and one considered Spark counts for more than twenty quick ones.
The AI analysis has its place here. It comes in as material the deliberation works with, not as the answer. Contributors ask what it means for this business, which assumption they would not trust, what it leaves out. That is the human checking McKinsey asks for, done by many people who know different parts of the business, rather than by one leader with a stack of reports.
If you are working out how AI analysis and human judgement should meet in your organisation, the Enterprise toolkit shows how Hunome runs a deliberation alongside the analysis you already have.
Alignment means knowing who is where, not making everyone agree
McKinsey is right that alignment building cannot be compressed. Today it is mostly done one of two ways. Rounds of meetings bring people round to a view, and end in apparent agreement that comes apart once the decision meets reality. Or a vote settles it, which does produce a decision, and throws the reasoning away at the moment of the count.
Hunome takes a different line. You do not need agreement to decide. You need understanding, and you need to know who is where and why. The Lens shows the person leading the deliberation where themes are forming and where tensions sit while the deliberation runs. The Shared Understanding Index reads how well contributors grasp one another's reasoning, disagreement included. Two people who disagree completely, and can each state the other's case, have high shared understanding. A room that nods along without ever making contact has none.
Whoever leads and decides, an executive team, a programme group or a board, then decides with that map in front of them. The accountability is theirs. The judgement behind it is wider than they could build alone.
Junior people need somewhere to learn judgement
McKinsey flags a quieter risk too. Junior staff used to build judgement by doing the analysis. If AI does that work and nothing replaces the experience, the next generation of leaders arrives with less of it. Training people to check AI output helps. It is not the same as learning how experienced people weigh a hard call.
A deliberation gives that learning a place. Junior contributors put their Sparks next to those of people with decades in the business. They see how a point is built on, challenged and connected to others, and they see their own thinking taken into account. Judgement is learnt by exercising it alongside others, and a SparkMap keeps that exercise open to everyone invited into it.
Faster review cycles need reasoning that stays
The authors note that faster analysis could let companies move from annual reviews to quarterly or monthly ones. That only helps if each review does not start from scratch. When conditions change, the useful question is not only what was decided. It is why, on which assumptions, and who saw it differently.
Deliberative intelligence is built for exactly this. It is shared understanding that holds, kept as a living structure of how a group is thinking, with the reasoning behind it preserved and available to be worked with later. When the next review comes round, the SparkMap is still there. The assumption someone was right to doubt can be found, with the reasoning around it. The deliberation picks up where the thinking left off, instead of where the slides did.
Widen the judgement, keep the accountability
The judgement bottleneck is real, and McKinsey is right not to wish it away. Judgement runs on human time, and squeezing that time is the wrong response. Rationing it among a few leaders is only half an answer.
The organisations that come out ahead will give that time to more of the right people, keep the reasoning those people build, and let the person accountable decide with the whole picture in view. Analysis can run on AI time. Judgement gets better when it runs on more people's time.
Source: McKinsey & Company, "The judgment bottleneck: Balancing AI and human speeds in strategy", McKinsey Quarterly, 2026. https://www.mckinsey.com/capabilities/strategy-and-corporate-finance/our-insights/the-judgment-bottleneck-balancing-ai-and-human-speeds-in-strategy
