Unconscious bias training tends to get commissioned after something has gone wrong. A comment lands badly in a meeting, a grievance is raised, or an exit interview surfaces a pattern nobody wanted to look at, and the request that follows is usually for a session to be booked as quickly as possible.
The session itself is rarely the problem. What matters is whether anything is measurably different six months later, and on that measure most unconscious bias training in the UK does not deliver. That has consequences beyond a wasted budget, because the standard UK employers are held to changes in October 2026, which makes the gap between training that is defensible and training that is merely documented considerably more expensive than it used to be.
What the Evidence on Unconscious Bias Training Actually Says
In 2018 the Equality and Human Rights Commission commissioned Doyin Atewologun, Tinu Cornish and Fatima Tresh to review the research on whether unconscious bias training works. Their report assessed the evidence against four separate aims: raising awareness, changing implicit bias, changing explicit bias, and changing behaviour. Keeping those four aims apart is what most summaries of the report fail to do.
On awareness, the training performs well, because people do come out of a decent session understanding how snap judgements form and where their own patterns show up.
On implicit bias, the picture is mixed. The EHRC review found that more sophisticated programmes can reduce measured unconscious bias for up to eight weeks after the intervention. What counted as sophisticated matters here, because those programmes combined three things: awareness of bias, genuine concern about its effects, and practical tools for countering it.
On behaviour, which is the outcome most employers care about, the evidence is thin. The reviewers noted that behaviour change is difficult to measure and that most studies never attempted it rigorously.
Two years later the Behavioural Insights Team reviewed similar territory for government and reached a blunter conclusion, finding no evidence that this type of training changes behaviour in the long term or improves workplace equality in terms of representation or pay gaps. In December 2020 the Cabinet Office announced that unconscious bias training would be phased out across the civil service in England, and encouraged other public sector employers to do the same.
Why the “It Doesn’t Work” Headline Is Only Half True
The difficulty with treating that announcement as the end of the argument is that it collapses a specific finding into a general one. What the research demonstrates is that a standalone awareness session, delivered once, disconnected from any change in how decisions get made, does not shift behaviour.
Bias in hiring and promotion did not stop existing because a government department stopped running slides about it. The useful question is no longer whether an organisation should talk about bias, but whether talking about it is the whole intervention or merely the opening ten minutes of one.
Is Unconscious Bias Training a Legal Requirement in the UK?
No. There is no statutory duty under the Equality Act 2010, or anywhere else in UK employment law, requiring an employer to run unconscious bias training specifically. The legal obligation is to prevent discrimination and harassment, and the method is left to the employer.
Where this becomes serious is in how a tribunal assesses whether the employer tried. Since 26 October 2024, section 40A of the Equality Act has placed a preventative duty on employers to take reasonable steps to prevent sexual harassment, shifting the burden from reacting to complaints towards anticipating risk. From October 2026 that standard rises again, and employers will need to demonstrate they took all reasonable steps, alongside new direct liability where a worker is harassed by a third party such as a client, contractor or member of the public.
Will a Training Session Protect You at Tribunal?
Only if it survives inspection, and the bar for that is rising. Under the current duty, many employers have relied on an annual e-learning module and a signed policy acknowledgement. From October 2026, tribunals are expected to look past whether a policy exists and into how it was implemented, including the frequency and the quality of the training behind it.
The volume of claims makes that scrutiny more likely. There were 64,000 open single employment tribunal claims at the end of March 2026, a 55 per cent rise on the previous year, while disability discrimination claims grew by 79 per cent over the same period to reach 4,056. Where the only evidence of preventative action is a completion report showing that staff clicked through a module, that will not read well by the time it is examined.
Why the Standard Half-Day Session Falls Flat
The first failure is that the session stops at awareness and leaves people there. Telling a group of managers that bias is universal, automatic and largely invisible, without giving them anything to do about it, is not neutral in its effect. It can quietly license the behaviour, because if everybody does it and nobody can help it, then nothing is anyone’s responsibility.
The second is that it happens once. Even the sophisticated programmes in the EHRC evidence held their effect for only around eight weeks, which leaves ten months of the year uncovered by an annual session, and very few programmes are designed with that in mind.
The third, and the one that shows up most often in senior teams, is that the training sits apart from the decisions where bias costs money. Nobody leaves the room and changes how a shortlist is built, how a calibration meeting is run, or who gets handed the visible project. The workshop happens in one place and the decisions happen in another, and the two never meet.
What Works Instead of a One-Off Session
The approaches with the most evidence behind them share one feature, which is that they treat bias as a property of processes rather than a property of people.
Start With the Decisions, Not the Attitudes
The judgement calls are the place to look first. Who sits on the panel, how the criteria are set, what happens in the room where performance ratings are moderated, who receives stretch work and who is given the reliable but invisible tasks. This is the substance of a proper culture audit, and it surfaces the real pattern faster than an engagement survey does, because processes hold the pattern more reliably than opinions do.
Build Cultural Intelligence, Not Guilt
Cultural intelligence gives leaders something to practise, which awareness on its own does not. It is the capability to work effectively across difference, whether that difference is geographic, generational, functional or identity based, and unlike an unconscious association it is learnable.
Anglie has delivered cultural intelligence workshops with executive teams at organisations including Atkins Realis and Nomad Foods. The emphasis there is practical rather than confessional, which is part of why those rooms tend to be less defensive.
Train Managers for the Moment, Not the Theory
The skill that matters is naming something in real time without damaging the relationship. That means rehearsing the difficult moment, the comment in the meeting, the repeated interruption, the “culture fit” objection that nobody can quite define. This sits much closer to leadership team training than to compliance, and it is the part managers consistently report they were never given.
Measure Something Real
Two or three decision points, tracked over time, will tell an organisation more than any post-session feedback form. Shortlist composition against the applicant pool, promotion rates by group, and who is put forward for high-visibility work are all straightforward to monitor. A shortlist that looks materially different twelve months later is also the kind of evidence that supports an all reasonable steps position.
How to Choose an Unconscious Bias Training Provider in the UK
For organisations commissioning this work, these questions are worth asking before anything is signed.
- What is this designed to change, and how will we know whether it happened?
- What happens after the session, and across what period?
- Which of our decision processes will this touch?
- How do you handle the participant who thinks the whole exercise is nonsense?
- What will you leave behind that a tribunal would accept as evidence of implementation?
A provider who answers the first question by describing their slide deck is selling a completed attendance list. That was defensible in 2019 and it will be considerably harder to defend after October 2026.
Where to Start
For organisations whose current provision is an annual e-learning module, the sensible sequence is to examine the decision processes first, establish where the risk sits, and design the development around what that reveals.
Anglie Johns works with organisations on exactly this, whether through a culture audit, a targeted programme with a leadership team, or stepping into the business directly when the situation calls for it. To work out what your organisation needs before the autumn, book a consultation.
