Work Transformers Labs · Research

Project Lighthouse

Applied research into the human side of AI: leadership, adoption, governance and how teams come to trust AI-supported work.

Applied researchOngoing programme
The state of play
The technology is no longer the hard part. The organisation is.

More than half of organisations describe their AI use as a chaotic free-for-all, four in five say it is being built in silos, and three in four admit governance has not kept pace with adoption. The models work. What is missing is leadership, trust and a way to govern the work.

55%
describe their AI use as a chaotic free-for-all
Enterprise surveys · 2026
79%
say AI is being built in silos
Enterprise surveys · 2026
3 in 4
say governance has not kept pace
Enterprise surveys · 2026
Shadow AI
The tools are already in the building.

Whether or not a firm has a strategy, its people are already using AI. Two thirds of executives believe their organisation has already suffered a data leak through an unapproved tool, and almost all accept that generative AI makes a breach more likely. Banning it does not work. Governing it does.

Ungoverned use

Staff adopt consumer tools quietly, outside any policy, permission or review.

Data exposure

Sensitive documents are pasted into tools with no control over where they go or how they are used.

Inconsistent quality

Without shared method or review, output quality varies by whoever happened to prompt it.

The research question

Project Lighthouse studies the human system around AI: how leaders sponsor it, how teams come to trust it, how it is governed, and how adoption actually takes hold in serious firms. The question is not whether the technology works. It is what has to be true for people to rely on it.

The trust gap
Adoption is a trust problem before it is a technology problem.

Younger staff already trust AI more than their managers for some tasks, yet only about a third say their manager is an AI champion, and skills and training gaps are the single biggest barrier to responsible adoption. The result is a gap between what people will use on their own and what leadership is ready to stand behind.

80%
of Gen Z trust AI more than their manager for some tasks
Enterprise surveys · 2026
35%
say their manager is an AI champion
Enterprise surveys · 2026
#1
barrier to adoption is skills and training gaps
Enterprise surveys · 2026
How it fits together
Adoption is built, not announced.

Start with the tools already in the building, earn trust through real use, and turn pilots into lasting change.

Existing toolsTeam feedbackUsage dataSystemsLeadershipHigher adoptionEarned trustDaily useFewer workaroundsConfident teamsLasting changePilotTrainBuild trustEmbedSupportMeasureChampionIterateScaleTHE STARTING POINTEARNING ADOPTIONWHAT CHANGES
What we are exploring
Six questions that decide whether AI is adopted.
Leadership behaviours

What sponsors actually do, in the open, that makes a team adopt rather than resist.

Adoption patterns

Which systems earn trust first, and how trust spreads from there.

Governance models

The policy, permissions and review that turn shadow use into sanctioned use.

Trust mechanisms

How showing reasoning, sources and confidence changes whether people rely on an output.

Workforce uplift

How juniors reach senior-quality output, and how senior time is freed for judgement.

Measurement

How to tell real adoption from activity, and value from usage.

What leaders do
Five moves that earn adoption.
01
Sponsor it visibly

Adoption follows leaders. Where a senior sponsor uses the system in the open, teams follow; where AI is delegated downward, it stalls.

02
Start where trust is cheap

Begin with high-volume, lower-stakes work, prove the output, then earn the right to the harder decisions.

03
Show the reasoning

Trust comes from seeing the working, the sources and the confidence, not from a black-box answer.

04
Govern in the open

Clear policy, permissions and review turn quiet shadow use into sanctioned use, and remove the fear that blocks adoption.

05
Measure uplift, not usage

Count better, faster, more consistent decisions, not logins or prompts. Usage is vanity; uplift is the point.

What clients can learn

Firms working with us through Lighthouse leave with an honest picture of their own adoption: where AI is already being used unofficially, where trust is missing, and what leadership, governance and change support it would take to make AI a normal part of how the work is done.

Where this is heading
The agentic era rewards the firms that earn trust, not the firms that move fastest.

As systems take on more of the work, trust and governance become the constraint on value, not model quality. Most leaders already believe responsible AI improves returns; far fewer have turned that belief into operating practice. Closing that gap, between principle and practice, is the work of the next few years.

Related research and outputs
How we publish

What we share, and what we keep.

Project Lighthouse is open research into the human side of AI: leadership, trust, governance and how adoption actually takes hold. We publish the questions we are wrestling with, the patterns that recur across real deployments, and the principles that hold up under pressure. We do not publish client data, or the parts of the method still being worked out. It is research, not a product: the questions are the interesting part, and the answers are earned in the work.

Work Transformers Labs

Following the adoption research?