This blog captures the key themes from Audacia's panel event, Managing Change with the Rapid Evolution of Technology, held as part of Leeds Digital Festival 2026.
Leaders from central government, health technology and the health, insurance and legal services sectors came together to discuss how organisations manage change when it no longer has an end point. The panel was chaired by Philip White, Founder and Managing Director at Audacia, with panellists Cleveland Henry, Interim COO at Enlivio; Mazella Heaton, Transformation Director at Kuro Health; and Paul Thomas, Deputy CTO at the Department for Business, Innovation, Science and Trade.
The views here reflect the collective discussion rather than any single individual or organisation.
The Pace of Change
Change has always been part of working in technology, but its speed and the expectations that come with it are new. Planning horizons have shrunk from years to months to days, and with AI tools now widely available, boards and colleagues expect results almost immediately.
The way large organisations deliver change has shifted too. Ten years ago, transformation typically meant a two-to-three-year programme with a defined end state. Now it is a continuous process, delivered in short cycles towards a longer-term goal. This way of working has historically been more typical of smaller organisations, where limited resources and funding have never allowed for long programmes.
For all that, the technology is rarely the hard part; people and processes decide whether change lands, and continuous change gives them no natural pause. In some software teams, four-week sprints have become 24-hour turnarounds, leaving little time to plan or reflect, and burnout is a growing consequence.
Pressure to Move Faster
Much of the pressure to move quickly comes from senior leaders responding to what they see in the market. Implementing change without due diligence leaves teams delivering work or adopting tools that add little, if any, value. Launching a customer-facing tool without time for user research or service design, for example, can result in a product nobody asked for, and nobody uses.
Demand also comes from the front line, where colleagues now regularly ask product teams to do things “agentically” with Copilot, often far beyond what the organisation can deliver, and explaining why takes up time.
The same urgency is shaping workforce decisions. AI was promised as a way to make things better, safer and faster, but in practice the conversation has narrowed to faster and cheaper. Some boards are already reducing headcount based on what they expect AI to deliver, before any benefit has been proven. The panel's view was that organisations which see the change through first, and reduce costs only once the gains are real, often achieve better results.
Some organisations are resisting this pressure and benefiting from it. In one example discussed, mental health clinicians now use AI scribe tools to record consultations instead of typing them up afterwards. Instead of using the time saved to see more patients, they spend longer with each one, delivering a more valuable service.
Risk Management Is Struggling to Keep Up
Governance and risk management are not keeping pace with technology change, and the gap is clearest where systems act on their own. Self-healing platforms that find, fix, test and deploy security patches without human input offer an obvious security gain, but the governance needed to oversee changes made without a person in the loop is still catching up. Leaders who have seen AI models hallucinate are right to question how far to trust these systems in the meantime.
At the same time, risk appetite is shifting towards making a good decision now over a perfect one in six months, though how far that goes depends on the sector. In banking, a mistake might mean unpaid mortgages; in health, the consequences could be more severe. One panellist described the right balance as “risk-appropriate innovation”.
Recovery raises a bigger concern still. For example, organisations recovering from the CrowdStrike outage were able to roll it back because human error had caused it and they understood what had happened. As more change is automated, that understanding can’t be assumed, and a major incident that nobody can trace or reverse becomes a realistic prospect.
Furthermore, the controls themselves have limits. If capable systems can find ways around their guardrails, organisations may need guardrails for those guardrails, and people to manage them. Risk registers can give similar false comfort, missing exposures nobody has considered, particularly when applications and teams move between organisations or onto new platforms.
For now, many organisations are moving carefully. Progress is often described as a path from adoption to augmentation to automation, and by that measure, many are still at the adoption stage. Where organisations are moving further, it tends to be in the back office, with process automation and single-function agents, while AI that affects customers' decisions remains the area of greatest caution.
The Cost of Self-Service AI
More business users want to build their own AI tools, often prompted by vendors marketing directly to them. It is the low-code debate returning in a more powerful form, with cost as the immediate concern. For example, in one organisation, an individual used 30 days' worth of tokens in two days. This is a particular consideration as newer, more capable models consume more.
There is a parallel with the wave of cloud migrations over the last decade, when many organisations moved legacy systems into the cloud largely unchanged, without redesigning them to run efficiently there. As a result, costs were often several times higher than expected. So far, there is no settled way to manage these costs beyond controlling who has access to which AI tools.
Measuring What Matters
Adoption figures say little on their own: a rise in Copilot usage from, say, 41% to 47%, matters only if the organisation knows what it is trying to achieve, whether that is throughput, time saved or something else, and checks whether it is happening.
This is where shorter delivery cycles help. With clear measures of value in place, continuous change makes it easier to spot work that isn’t delivering and change direction early.
Supporting the People Who Deliver Change
As change becomes constant, people need more support, and that starts with communication. Leaders need to explain why the organisation is changing, what it expects to gain and what it means for individuals. Without that, people assume the worst, most of all when automation puts roles in question.
It also means protecting teams, including managing expectations upwards and being willing to say no when teams are asked to do more with less. The same applies to security, where people are often the target: staff need to take training seriously without feeling the whole responsibility rests on them.
One example was given of an organisation testing a more sustainable model, building slack into operations so teams aren’t running at full capacity and running change through multifunctional squads with staggered start dates. Each squad gets the last two weeks of every cycle off before the next one begins.
The same thinking applies to bringing new people in, because as tools write more of the code, curiosity, critical thinking and empathy matter more than any specific technical skill, and attitude and an appetite to learn count for more than a degree.
That raises a concern about the entry point into the industry. If organisations stop hiring junior developers because tools handle the basics, it is unclear how the next generation builds the foundations that senior roles depend on.
Final thoughts
The principles for managing change are familiar - understand the value before acting on it, know what is being changed and how to reverse it, and look after the people doing the work. Following them consistently is harder when change never stops, which is why they matter more now than when change came in programmes with an end date.
