
Since January, I have been leading the initiative to embed AI (artificial intelligence) across my company. It has been a significant undertaking. While there is still much to do, a clearer picture of the next steps is beginning to emerge. I wanted to share some of our key learnings to date, particularly for organisations looking to move beyond AI experimentation and integrate it into business-as-usual operations in ways that deliver tangible business value.Â
Each step has been important in terms of organisational learning and best practice:Â
Production use cases matterÂ
You need clear targets. We therefore require every department in the company to deliver at least one AI innovation per quarter, giving an annual company-wide target of 25 AI systems in production for 2026–27.Â
We already have 20 use cases in production and are on track to reach our target of 25.Â
I set these goals because you need something tangible to work towards. Our experience in enterprise IT has shown that the value of a solution only starts to become measurable once it is in production.Â
I’ve also cautioned both leadership and colleagues against focusing exclusively on outcome KPIs. Yes, ideally, we should focus most of our time and energy on projects that drive clear financial ROI. But education is also an important part of a transformation of this scale. It’s also important to just get people to think and talk about how to do things differently and more efficiently with the use of new technology, and encourage them to challenge the status quo. [Text Wrapping Break][Text Wrapping Break]And if you dismiss each new idea because there is no clear cost benefit, you risk stifling creativity and innovation. By giving ideas room to develop, more ideas come.Â
AI Adoption Tip #1: Set targets for getting AI into production, not just for business impact. You can’t properly measure the value of an individual project without getting it into production first. Don’t underestimate the value of getting the ball rolling!Â
Give AI a friendly internal face Â
Right from the start I told the Board that AI won’t be adopted if people are frightened of it. One of our key initiatives has therefore been Botsi, a flagship internal agent that integrates with the different tools in our ecosystem and provides employees with a familiar, accessible AI interface for finding information. The aim is to make accessing and working with information as straightforward as possible. Whether employees need information from Salesforce, Jira, Confluence or other systems, they can access it directly through the agent in Microsoft Teams.Â
I think that was a very useful foundation for our wider AI adoption efforts. We are continuing to build new digital use cases into Botsi to maintain momentum. These range from automatically updating Salesforce using call transcripts with prospects and customers to a growing range of opportunities to deploy agents to accelerate processes and make everyday work more efficient.Â
AI Adoption Tip #2: Don’t overwhelm people with 100 different agents. Give them one familiar, reliable front door they’ll want to keep coming back to.Â
Bottom up progress Â
I went into this role determined that AI would only become part of everyday working life and a driver of change if it was led by the people on the front line, rather than imposed from above. That means encouraging every department to come up with what they see as their biggest inefficiencies per department. We then work with them to prioritise the areas where AI could deliver the greatest improvements, before bringing in the technical team to help implement the solutions.Â
This approach doesn’t just keep the AI train moving; it’s also a vital part of the change management challenge that introducing AI inevitably brings.Â
Training is at the heart of this bottom-up cultivation of AI enthusiasm. We use training resources from major vendors like OpenAI and Claude, but I’ve also developed some of our own training. To be honest, you need both a carrot and a stick. The carrot is showing people how AI can make their working lives better and inspiring them to explore the possibilities. The stick is making it clear that, as a company, we’re moving in this direction and expect people to use these tools and develop the necessary skills.[Text Wrapping Break]Â
AI Adoption Tip #3: Actively work on increasing the average knowledge level on the technology. Get people to work with the technology hands-onÂ
Measure for measureÂ
From the outset, I also made sure we had mechanisms in place to track adoption and understand how employees were responding to AI. Metrics and surveys are therefore a big part of my team’s work. We do a quarterly employee survey where we ask questions about their experience of the organization including how often they use AI. We combine this feedback with usage data from our AI tools to build a picture of how far AI is genuinely becoming part of people’s day-to-day workÂ
Other benchmarks suggested that 50% adoption was a good target. But after our first six months, we were already well ahead of that, at around 69%. So I raised the bar, setting a new target of 75%, well above the benchmark for AI usage in our industry. I’m delighted that we have achieved and exceeded that level of adoption. Within that, about 50% of our people are heavy users, while 30% are moderate users. Another 20% lag behind, at least for now.Â
Would I prefer 100% on all of these metrics? Of course. But I think we’ve made very good progress.[Text Wrapping Break]Â
AI Adoption Tip #4: Don’t use vanity metrics, use real frontline engagement data. And keep checkingÂ
You also need to make AI adoption as micro as possible, looking for opportunities to speed things up everywhere, however small. For example, creating automatic meeting summaries based on call transcripts, and potentially updating third party systems like salesforce.Â
Another example is our customer support knowledge base. By using agents to mine historic data, we have been able to build out our knowledge base to a level way beyond what was previously possible. This enables customers to find answers themselves or have support tickets answered automatically, without human intervention.Â
We use AI coding agents heavily in the organisation and are already seeing >50% productivity increases. We are now taking the next step into a full cycle software factory, where agents are able to run the entire software development lifecycle. We expect this to double productivity in engineering.Â
AI Adoption Tip #5: Encourage an ‘Is there an AI workaround for that?’ mindset wherever you can.Â
AI outcomesÂ
It’s still early days for AI as business as usual here, but I can already confirm that people are making savings. AI impacts many different areas of work, although some have more impact than others, and some are easier to measure versus others. We see the biggest impact in engineering, where productivity has increased by 50%. Outside of engineering, people are saving approximately 20% of their time, and that figure is continuing to grow.Â
But just as important as these benefits is the need to minimise risk. We’ve built AI governance and guardrails into our approach from the outset. Nothing goes forward to rollout from ideation here without Legal and IT carrying out due diligence on possible implications from a new idea around data privacy, security, and so on. We also have strict rules on where AI can and cannot be used. It cannot be used for reviewing employee performance or determining salary, for example. We’ve built in triggers so that if someone attempts to use AI in ways that breach company policy, our agents will flag the issue and block the action or require additional checks.Â
AI Adoption Tip #6: Look for AI benefits, but always within clear, responsible boundaries.
AI can deliver value from day one, provided you manage the gap between what the technology can do and what your people are ready to do with it.Â
Don’t wait for the perfect enterprise-wide strategy. Start with real problems, put solutions into production and measure the results. Train and inspire your people, build in governance, and embed AI into everyday operations to help your company work not just differently, but better.Â
The author is Chief AI Officer (CAIO) at global leader in document intelligence, DoxisÂ
Â



