June 24, 2026 · 5 min read

The missing role in AI transformation.

Most organisations have owners for the work. Few have owners for the connections between the work — and the connections are where AI initiatives stall.

The demo always works now.

That is the strange part. The model performs, the pilot proves the concept, and then almost nothing reaches the people it was built for. We have made the technology easy and the outcome is still rare.

The conversation around AI still focuses on capability.

  • Better models.
  • Better agents.
  • Better tooling.

Yet most organisations already have capable technology.

What they struggle with is turning it into value.

The bottleneck is not intelligence. It is a meeting that never quite happens.

A pattern I keep seeing.

01A team identifies a promising use case
02A prototype proves the concept
03Funding appears
04Stakeholders get excited
05Then progress slows

Fig. 1 — The demo works. The momentum does not survive the movement through the organisation.

The part that works The model works. Capability, tooling, the prototype
The part that doesn’t The organisation doesn’t. Ownership, coordination, trust

Fig. 2 — The failures are less often in the technology. They are in everything the technology has to pass through.

The initiative now crosses multiple boundaries.

The initiative
01Data ownership
02Risk
03Compliance
04Operations
05Customer experience
06Workforce capability
07Budget
08Governance
Whether value gets through

Fig. 3 — Each boundary has its own owner. Value has to pass through all of them.

Each group arrives with different objectives, constraints, and incentives. The challenge is not persuading people to use AI; it is getting groups with different priorities to decide together.

That requires a role that rarely appears on an organisation chart.

The Bridger

Most organisations already have one. Few of them know it, and fewer still reward it.

Harvard Business Review research on why great innovations fail to scale found that successful bridgers consistently do three things: they curate, translate, and integrate.

Job 01CurateBring the right people into the room — including the ones no one thought to invite.
Job 02TranslateTurn each group’s vocabulary into a problem the others can recognise.
Job 03IntegrateBuild shared understanding of success and the trade-offs everyone can accept.

The first job is curation.

Most AI programmes bring together the obvious people. Technology. Product. A sponsor. Perhaps legal or risk later.

The bridger thinks differently.

01Who can accelerate the work?
02Who can block it?
03Who understands the operational reality?
04Who will inherit the consequences?

Many AI initiatives fail because the people who determine success were never in the room.

The second job is translation.

One of the most common sources of friction in organisations is not disagreement.

It is vocabulary.

Engineering talks aboutData quality
Operations talks aboutProcess consistency
Product talks aboutTrust
Users talk aboutReliability
All four describeThe same problem, viewed from different positions.

Fig. 4 — Four teams, four words, one problem underneath.

A surprising amount of organisational conflict disappears when people realise they are talking about the same thing.

The third job is integration.

Most organisations are good at creating activity. Meetings. Workstreams. Governance forums. Status updates.

Activity is not alignment.

Integration means creating shared understanding of what success looks like and what trade-offs are acceptable.

It means surfacing assumptions before they become politics. It means helping people understand not only what others are doing, but why.

That is where trust starts.

Effective bridgers pay attention to fear.

People protect what they care about.

01Quality
02Autonomy
03Customers
04Expertise
05Reputation

Fig. 5 — When someone pushes back on an AI initiative, the resistance is usually protecting one of these.

So the interesting question is not:

The obvious question “Why are they resisting?”
The better question “What are they protecting?”

Fig. 6 — The answer reveals something important about the system.

What decides success is whether enough mutual trust, influence, and commitment exist for people to move together.

Less exciting than agents, copilots, and autonomous systems.

It decides whether any of them matter.

As AI capability continues to improve, the bottleneck will move elsewhere.

Coordination
Trust
Organisational learning
Working across boundaries

Fig. 7 — The same scattered owners from Fig. 3, now connected. This is the work that rarely shows up on a roadmap.

So the field keeps shipping better models, better agents, better tooling. In only a few cases is that still the bottleneck.

The demo was never the hard part. The hard part is the role that carries it across the boundaries — the one that never shows up on the org chart.

Succeeding here has little to do with model choice. It comes down to connecting capability, adoption, and change — because the connections decide whether value gets through.

If you have a bridger in your organisation, say so — out loud, to their face, this week. They are usually the reason good ideas survive contact with reality.

Andreas Conradi · June 2026