Change is a human challenge
In the Civil Service, organisational change is a reality. Teams evolve, priorities shift, and departments periodically go through restructures, machinery of government changes, and workforce redesign programmes. While each change programme is different, they all have one thing in common: they affect people.
Change is ultimately a human challenge. It impacts careers, wellbeing, and confidence in the organisation. Technology cannot solve those challenges, but it can support the processes around them by making them more transparent, consistent and fair.
I became involved in a large organisational restructure because the allocation process presented exactly that kind of opportunity. The restructure involved allocating a large number of people across multiple grades, limited vacancies and competing preferences, all within a complex set of agreed business rules. Managing this fairly and consistently through a manual process was increasingly difficult. The challenge was not simply to make the process faster, but to reduce risk and give people greater confidence in the outcome.
The experience reinforced an important lesson: when technology is used in people-focused decisions, transparency and accountability must come before sophistication.
The challenge of organisational restructureThe existing allocation process involved significant manual effort. Candidates could express preferences across multiple roles, vacancies were limited, and allocations needed to consider scores, preferences and redeployment rules simultaneously.
Managing this process manually created several challenges, including:
large volumes of candidate and vacancy preferences multiple interdependent variables affecting each allocation significant time required to reach outcomes greater risk of human error when processing large datasets difficulty providing clear explanations of how individual outcomes were reachedMost importantly, long processing times meant people spent longer waiting for clarity about their future roles.
The objective was therefore not simply operational efficiency. It was to reduce the risks associated with complex manual processing to improve outcomes for the people involved.
Using technology to support (not replace) decision makingAt the heart of the solution was an allocation algorithm that matched candidates to roles.
The word "algorithm" can sometimes imply automated decision-making. In this case, the system acted more like a transparent administrative process than an AI model.
The algorithm was “deterministic”, which meant it followed a defined sequence of rules:
Rank candidates consistently according to their interview and written statement scores (which were assigned by people). If ties, rank candidates on the vacancy holder’s preference, and then the candidates’ preference scores. Allocate roles according to predefined business rules. Re-run the algorithm automatically if there are still ties or un-allocated candidates and roles. If no further allocations are made in a run, the re-running stops. Escalate unresolved cases for human review.Importantly, the system did not learn, predict or optimise like an AI model. It simply applied the agreed rules consistently and at scale. Every allocation could be traced back to documented rules and independently verified.
Choosing automation over artificial intelligence was a deliberate choice to ensure:
the same inputs always produced the same outputs every outcome could be fully explained every decision path could be audited the process remained transparent and therefore legally defensible human decision-makers retained accountability to make judgements and approvals Ethics and governanceGovernance and ethical oversight were built into the project from the outset.
Alongside HR and trade union engagement, we worked with the departmental ethics committee to review the proposed approach.
The allocation process was never intended to replace human judgement. Cases that could not be resolved automatically were escalated for review. Decision-makers retained the ability to challenge or override outcomes where there was a clear and documented rationale.
This early engagement helped shape both the design and the governance framework, while keeping stakeholders informed of the benefits of the approach prior to implementation.
Testing and assuranceBefore deployment, we ran the allocation process in parallel with the existing manual method.
This allowed us to:
validate allocation outcomes verify that business rules were being applied correctly identify any discrepancies measure efficiency gains build confidence among stakeholdersThe comparison demonstrated that the automated approach produced consistent and explainable outcomes while dramatically reducing processing time.
ImpactThe system reduced a process that previously took days to complete to a matter of seconds. But its most important benefits were not technical.
For the people involved, this meant less time waiting for clarity during an uncertain period. For decision-makers, it created a transparent, auditable record of every allocation, reduced the working hours needed to complete the process, and made each decision easier to explain.
“An important goal of the Business Growth Transformation process was to do all we could to reduce stress for those directly affected by the changes. This included not just simplifying the application process, but ensuring the decision-making process was clear, consistent and quick. Working in partnership with the Data Analysis Team in DDaT helped us achieve these aims.”
Ros Lynch, Deputy Director for Creative and Education Services
The manual method used for comparison produced several errors that the algorithm did not. There was only one instance where the human reviewer overrode the algorithm’s output: allocating someone their second choice freed up their first-choice role for an unallocated candidate.
The experience shows that government organisations do not always need advanced AI to solve complex problems.
What’s next?There are clear lessons for organisational change across government.
Many workforce and organisational challenges involve large volumes of information, complex business rules and significant consequences for individuals. These are exactly the situations where well-designed automation can add value.
The approach is also highly replicable. Because the solution relies on documented business rules rather than specialist AI models, it can be adapted by other departments facing similar allocation or prioritisation challenges.
Successful change is about improving outcomes for people. This can be supported with technology by involving users and ethics committees from the beginning.
Conclusion
If there is one takeaway from this work, it is that technology should make important decisions easier to understand, not harder. You do not always need AI to solve complex public-sector problems. Well-designed automation can support decision making while remaining understandable and auditable.
For organisations that feel cautious about using technology in sensitive decisions, the lesson is simple: start with explainability. People’s confidence in your solution will grow when they can understand how the system works, challenge its outputs and see that human judgement remains part of the process.
If you would like to know more about the algorithm or would like advice on replicating our work, please reach out in the comments.
seen at 16:46, 12 August in Digital trade.