UK Cross-Agency Collaboration - How to Make It Work

In UK public sector work, cross agency collaboration is usually the difference between a policy that looks tidy on paper and one people actually experience as joined up. It matters whenever housing, health, justice, finance, data, or local delivery overlap, because no single body owns the whole problem. In this article I explain what that work really means, when it is worth the coordination cost, which delivery models are used in practice, and how leaders keep accountability intact.

What cross-agency working means in UK government

I usually start with the simplest distinction: coordination is not the same as collaboration. The NAO’s good practice guide defines cross-government working as more than one department or arm’s-length body working together to achieve a shared outcome, which is a useful standard because it forces clarity about purpose, not just goodwill. In practice, the work can range from sharing data and lessons learned to running a joint programme with common governance, but the key test is whether the organisations can point to one outcome and one version of success.

This matters because the UK public sector is built around departments with their own mandates, budgets, and accountabilities. That structure is efficient for ownership, but awkward for issues such as net zero, rough sleeping, vulnerable families, or adult social care, where decisions in one place affect delivery in another. Once you see the problem that way, the real question becomes not whether to collaborate, but how much joint working the issue actually needs.

That distinction matters, because the decision to collaborate should come before any discussion of structure.

When the extra coordination is worth it

There are a few signs that the extra coordination is justified. I look for shared clients, overlapping responsibilities, a common geography, a single data problem, or a policy that will simply push work from one body to another. If the public would otherwise receive conflicting messages, duplicate assessments, or repeated requests for the same information, joint working is probably cheaper than trying to fix the fragmentation later.

• One service user journey crosses several agencies.
• One decision changes cost, risk, or workload elsewhere.
• One dataset or eligibility rule sits in multiple hands.
• One team cannot deliver without another team’s authority or expertise.
• One public priority would lose coherence if each body worked separately.

I am also cautious about overusing collaboration. It adds meetings, negotiation, and sometimes slower early decisions. If one body can lead cleanly and others only need to be consulted, I prefer a lighter model. The point is not to make every problem collaborative; it is to match the structure to the shape of the problem.

Once that is clear, the next step is choosing the right operating model instead of defaulting to the most complex one.

The six operating models used in practice

HM Treasury’s Managing Public Money sets out six joint working models, and that list is useful because it shows collaboration is not a single pattern. In my view, the best model is usually the lightest one that still gives you enough authority, visibility, and control. A shared-service arrangement, for example, may be perfect for IT, HR, or finance, while a joint programme with its own governance makes more sense when the policy is truly cross-cutting.

Model	Best fit	What it looks like	Main risk
Collaboration	Shared policy development	Departments work together on a policy where each has a legitimate interest.	Goodwill without clear decision rights.
One lead department with expert support	Need for specialist input but a single owner	The lead body formally draws on analysis, expertise, or assurance from others.	The lead can over-rely on advice and still own the wrong answer.
Separate funding for separate elements	One shared outcome made up of distinct workstreams	Each department funds its own contribution to the wider plan.	No one feels accountable for the whole.
Programme-level lead	Large cross-cutting programmes	An overall SRO oversees the programme while individual projects keep their own accountability.	Too much reporting, too little decision-making.
Budget cover transfer	One department wants another to deliver a common aim	Money is transferred so the receiving body can carry out aligned activity.	Funding moves faster than governance.
Machinery of government change	Policy responsibility itself needs to move	Formal transfer of responsibility, not just closer working.	Used when structure should change, not merely behaviour.

That table is not a hierarchy in the moral sense. It is a practical ladder. I would start as low as possible and only move up when the problem genuinely needs stronger integration, because every step up adds coordination cost and oversight burden.

The choice of model is only half the job, though, because even the right model fails if the governance is loose.

Governance is where most collaborations win or fail

This is where I slow teams down on purpose. I want the lead body, the senior responsible owner, the finance lead, and the operational leads to agree what decisions they can take alone and what must go through a joint board. If that sounds bureaucratic, it is actually the opposite: clear decision rights reduce the number of issues that end up bouncing between desks.

HM Treasury’s Managing Public Money is explicit that joint activity needs clear responsibility, including a single accounting officer, and that point should not be treated as a technicality. It is the difference between a shared programme and a shared misunderstanding.

• A short memorandum of understanding that states the outcome, scope, and roles.
• A named lead department and a named senior responsible owner.
• A shared risk register with agreed escalation thresholds.
• A finance map showing who funds what, when, and under which authority.
• A data-sharing plan that covers lawful basis, quality checks, and refresh cycles.
• A decision log so everyone can see what was agreed and why.

I also look for common data sets early. If two bodies need the same operational information, define the minimum shared dataset before the programme grows around assumptions. The cost of getting this wrong is not just technical rework; it is delayed delivery, duplicate reporting, and arguments about whose version of the truth is correct. Once governance is set, the real challenge becomes the human side of collaboration.

The leadership behaviours that keep the work moving

The strongest cross-agency leaders I have seen do not try to own everything. They act more like brokers: they keep the shared outcome visible, translate between organisational languages, and remove friction before it hardens into drift. That style is closer to facilitative leadership than command-and-control, which matters because collaboration rarely fails for lack of enthusiasm; it fails when nobody bridges the gaps between incentives, cultures, and decision cycles.

• State the shared outcome in one sentence and repeat it often.
• Make trade-offs explicit instead of hiding them in process.
• Use joint performance measures, not only departmental ones.
• Keep a regular rhythm of senior review so disputes do not pile up.
• Encourage boundary spanners, the people who can translate between teams and maintain trust across organisational lines.
• Reward co-operation publicly, because people notice where attention and credit actually go.

In service-heavy areas, co-location, joint training, or even a shared case conference can help, but only if the working relationship is mature enough to support honest challenge. The point is not to blend organisations into one blur; it is to let different bodies stay distinct while behaving as part of the same system. When that does not happen, the failure patterns are usually very predictable.

Common failure points I watch for

I see the same mistakes recur. The first is setting a shared goal but keeping separate incentives, so everyone agrees in principle and nobody changes behaviour. The second is building too much governance, which creates theatre rather than decisions. The third is treating data sharing as a late-stage admin task, when it should shape the design from the beginning.

1. Separate incentives, shared rhetoric - fix this by aligning measures and, where possible, funding.
2. Too many forums - fix this by giving one group real authority and clear escalation routes.
3. Unclear ownership - fix this by naming the lead body and SRO on day one.
4. Overengineering - fix this by choosing the lightest viable model.
5. One-off launch energy - fix this by building review points and evaluation into the operating rhythm.

The other failure I worry about is the quiet one: teams assume relationships will hold on their own after the first agreement is signed. They usually do not. Trust needs repetition, contact, and visible follow-through, especially when the bodies involved have different cultures or funding pressures. That is why the final thing I would do is decide what to lock in first.

What I would prioritise first in a new collaboration

If I were setting up a new cross-agency programme in the UK public sector, I would start with six questions: what is the shared outcome, who owns the decisions, which delivery model fits the risk, what data must move, who pays, and how will success be reviewed after launch. If any one of those answers is vague, the programme is not ready for scale yet.

• Agree the outcome in plain English before discussing structure.
• Choose the smallest governance model that still protects accountability.
• Write down the minimum dataset and the rules for using it.
• Set 3 to 5 joint measures that show whether the system is working.
• Plan a review after the first few months, not just at the end.

That is why I treat cross-agency work as an operating model, not a slogan: when the outcome, authority, and information flow are designed properly, collaboration becomes manageable, even under pressure. When they are not, even capable teams spend too much time negotiating the basics instead of delivering for the public.