From idea to implementation: driving value through AI solutions
From ‘idea to implementation’ is a strange phrase for any product development, but particularly where it comes to AI. It’s easy to get caught up in the potential of AI and there’s nothing wrong with experimentation for inspiration but when we get into the realms of implementation, the point is to solve a problem and deliver value.
So, you’ve done some interesting discovery work, identified user needs and a problem worth solving. How do you go from that to delivering an AI solution and reaping the benefits at scale?
Start with the value
When Transform launched our AI Lab with the aim of helping organisations use AI to solve real problems, we leant heavily on the same old product principles we’d apply to anything else - start with value.
In the process of defining the value, we should question whether we’re solving the right problem. AI is really good at automating high volume, repeatable tasks for example. Many organisations will jump at the opportunity to try out this technology for efficiency. Don’t! Or rather, make sure it’s the right problem to solve in the first place. Does that process need to be in place? What’s it trying to accomplish and is there a better way to achieve a similar outcome?
AI is incredibly expensive to run- societally, financially and environmentally - so investing in technology that isn’t delivering value is irresponsible.
You’ll also need to be able to prove value, in the same way you would for any other product implementation.
As part of our methodology, we establish baselines and metrics in a Discovery/Define phase to inform any problem statements and enable ROI calculations later. For a recent project with the Department for Education for example, we looked at how we could scale the operations of a human-led service to more schools.
Culture is still eating strategy for breakfast
Referencing a brilliant blogpost from Emma Stace when we worked together at the DfE; culture is an essential building block to any long-term success.
Organisations that have been successful in going from idea to implementation have developed a culture where learning, trying new things, being data-driven and having the capabilities to iterate in rapid cycles, is valued.
Digital/data teams also need to be closely connected and aligned with partners in the wider business who are responsible for the data and the problem spaces. The best implementations are led by operational teams who can also lead colleagues through change and engage them in new ways of working and designing the solutions.
Experiment, experiment, experiment
Being evidence-driven is essential and this means rapid feedback cycles and experiments.
The cycle of implementing AI products necessitates learning from proofs. If we take an AI project lifecycle, we see learning happening at each stage – from model training through to implementation. Organisations need to be comfortable with experimentation to deliver AI solutions, which involves rapid feedback loops, being led by data and being able to clearly articulate a problem and hypothesis.
Experimentation also comes with failure, so organisations need to have the safeguards in place to protect users and operators and create ‘safe to fail’ (not risk-free) environments.
Think about operations early
When looking at an AI implementation, we’re not talking about a technology implementation (as with any product), we’re talking about a whole operation, often transforming or augmenting human operations. It’s important then to define how the product will interact with the whole service, and the role humans will play in it.
Team ownership
Having led ‘Lab’ teams at MoJ, DfE and now Transform, I believe there’s a role to incubate thinking and encourage rapid innovation, but ultimately problem space owners need to also own the backlog and have the tools, capabilities and experience to deliver on it. In short, don’t create separate AI teams or AI product managers -this is just hype.
Ethics and standards
The Government Service Standard is the minimum anyone needs to consider when developing a product or service in the public sector. Many private sector organisations will have something similar. AI solutions need to bake these standards in from the start. All of your thinking needs to happen with an ethical approach and in a standards-driven way.
We’re also increasingly aware of the wider impacts of the decisions we make, particularly where it pertains to AI that have a significant impact on the planet. Digital teams generally have high agency to make decisions about what they invest in.
If you want to hear more about the recent AI Lab we ran with the Department for Education or you’d like to talk about a free AI Lab offer, join our roundtable discussions at Government Transformation week on Wednesday 6th of November. Or reach out directly to Jack on jack.collier@transformuk.com.