It is easy to be seduced by the accessibility of Generative AI to jump right in, but it’s worth stepping back a little. A hasty approach can waste money, introduce risks, and store problems for the future.

1.    Spend some time building an understanding of what AI can do, what the different kinds of AI are and the types of AI uses other organisations have implemented. Ideally, your main sources of information on this should not be  the AI labs, your consultants or anyone else actively engaged in trying to sell you AI products or services. 

2.    Consider the most important problems you have to solve and then think how AI might be useful to address them. Starting with a broader objective of ‘getting value from AI’ is starting with means rather than ends and is likely to waste time and money.

3.    Spend some time and, where required, money, raising the standard of AI understanding within your organisation.  This has a number of benefits; firstly, much of the best understanding of where AI might be useful is likely to come from within the organisation, rather than top down.  Bringing your people in early also increases the chances of them engaging with AI related change. Additionally, it reduces the likelihood that people will see AI as a threat and, actively or passively, sabotage or undermine it.

4.    Put someone senior in overall charge of AI implementation. Successful implementation will require the co-operation of many parts of the organisation, and it should be driven by larger business and strategic goals rather than technical criteria.

5.    Get an early and thorough assessment of of the extent, nature and quality of your data. This may require expert assistance but will save a lot of trouble later on.

6.    Prepare yourself and your organisation for where the costs and effort will be required. A broad but non-intuitive consensus in the consultancy world is summarised by the Boston Consultancy Group (BCG) formula of 10-20-70, representing a focus that is 10% on algorithms, 20% on technology and data and 70% on people and processes. Overall, it can do no harm to to treat the exercise as one of business change rather than an IT implementation.

7.    Set some clear rules around the introduction of AI into processes, to ensure that risks are understood and managed, that successes are scalable and that there are clear criteria to evaluate successes and discontinue projects that don’t bring clear value.

8.    It is not a failure to decide that you have no projects that have an acceptable return on the risk, effort and cost required. Or even that you are content just to authorise some ChatGPT accounts for particular activities or that what looked like AI opportunities are more easily solved with more conventional, and well tried, automation.

[Mail for industry-specific checklists - list these - or get in touch for a free initial consultation]