The first step toward an AI strategy: good stakeholder interviews
An economist would call Generative AI (GenAI) a general purpose technology because - rather than solving some specific problem within a particular industry - it can be applied widely and creatively. GenAI new, powerful and evolving fast. We are confident that there are many more potential applications than have been demonstrated thus far. In fact, sometimes the only barrier preventing us from applying GenAI to a business problem is a simple lack of imagination!
An AI Strategy connects our vision for the future to the reality of today by identifying the best path from the latter to the former.
Developing an AI Strategy is a matter of synthesizing two modes of thought: the creative, exploratory mode where we search for opportunities, and the rigorous, logical mode where we critically evaluate and prioritize. In developing a successful strategy, you must consciously switch between these modes of thought throughout the process.
At Veratai, we follow a simple but well-refined recipe for developing AI Strategies called Discover. It consists of three tasks: Stakeholder Interviews, Research & Ideation and Analysis & Synthesis. (The first two of these can - and should - be run in parallel.)
This article is the first of three outlining our methodology. (The second in the series, Research & Ideation can be found here and the third, Analysis & Syntheis is here.) In it, we cover the Stakeholder Interviews task. We explain why and how we do it and offer some tips & tricks. Hopefully, by the end, you will be confident enough to successfully conduct interviews yourself.
Why do we do stakeholder interviews?
The purposes are threefold:
To discover the best opportunities for applying GenAI and to assess the benefits that would follow.
To uncover any blockers or dependencies that will need to be addressed before a solution can be implemented.
To build rapport with key stakeholders across the business and get a sense of how easy or difficult it will be to effect change.
Note that, at this stage we are not thinking in detail about how GenAI technologies would be applied. We save those questions for the next stage of the process (Research & Ideation). Here, we are purely focused on the opportunities, the risks and roadblocks, and the potential benefits. In fact, GenAI itself does not even need to be discussed in depth during the interviews; we ourselves have done many where it was never even mentioned!
Approach the interview process with an open mind as to where the conversations will take you. Certainly, anchor the discussion according to a defined scope or remit where you can – this could be a business strategy or a briefing document – but be willing to search wide and dig deep to find the less obvious opportunities. The keys to success are who you interview and how you prepare.
Who should you interview?
We normally conduct 6 – 10 interviews – but you could do more. The list of interviewees will vary, business to business, but it should be broad. Typically, it would include:
A member of the leadership team, who will give you a good idea as to how investment decisions are taken, how the business strategy is being implemented and what views exist within the board about GenAI. If you can, schedule these interviews first as they will often help to direct your focus during subsequent conversations.
A member of the sales team, to discover how the market views the business’s products and services as well as what competitors are doing and what customers are asking for.
Members of customer-facing support and “operational” teams. Employees who regularly interact with customers are a valuable source of information on how products and services are actually used, and where the most common problems arise.
Product managers and product developers. Where there is a digital component to the products and services offered by the organization, these employees will often have great ideas as to how they could be integrating GenAI. Furthermore, there are often opportunities to provide better data – or better insights from existing data – to these teams, helping them design the next iteration of products and services.
Members of the data, IT and analytics teams will be able to provide a good overview of the “data landscape” – what is available, what are the shortcomings. What data or insights requests can typically be fulfilled and what cannot? If you can, schedule these interviews last. By the time you speak to the data, IT and analytics teams, you should have a good overview as to the needs of other teams and will be able to ask specific questions around what data are available, discuss known quality issues and learn about any relevant infrastructure constraints that may exist.
Prepare for the interviews
A good Discovery interview is an exploratory process. At the outset, you don’t know what there is to learn and so it is important to prepare with this in mind. We find that the following structure strikes a good balance between open-ended exploration and detailed fact-finding. It’s helpful to think about three kinds of questions you can ask:
Open Questions: allow for broad, detailed responses. They typically begin with "how," "why," or "what" and encourage the interviewee to share their thoughts, experiences, or opinions in depth.
Semi-Open Questions: provide some structure but still allow for elaboration. You might offer specific options or a framework for the response but you’ll still leave room for the interviewee to expand on their answers. Semi-open questions are helpful for gathering detail or for impact-assessing ideas or problems.
Closed Questions: are specific and typically elicit short, factual responses, often "yes" or "no" answers or a choice from provided options. They are useful for gathering clear, concise information and are best used for gathering specific facts or for confirming your understanding.
When we prepare for AI Strategy interviews, we begin by jotting down a series of open questions relevant to the interviewee’s role. For a 45 minute conversation, you should need no more than 10 of these. It is important to presume as little as possible about the challenges you think the interviewee might face or the role that GenAI could play in their working life. Open questions are a great opportunity for the interviewee to share their expertise and understanding so don’t be afraid to range widely! Some of our favourites are variations on the following:
“What are the most rewarding / difficult aspects of your role?”
“Tell me about the last time you had to hunt for data to help you complete a task.”
“What was the last piece of feedback you received from a customer?”
“If you could re-allocate an hour each day from one aspect of your job to another, what would you stop doing and what would you do more of?”
“How has your role / industry changed in the last three years?” (“How do you think it will change in the future?”)
don’t control, Follow
You don’t need to ask all of your pre-prepared open questions; see them instead as springboards to get the conversation going. As the interviewee talks, you are listening out for promising avenues of enquiry which you can begin do dig into using semi-open questions. It will be helpful to keep at the forefront of your mind some of the situations where GenAI technologies are known to excel:
Time-consuming business processes that involve data from multiple systems.
Research and knowledge work that involves gathering and analyzing unstructured or semi-structured information.
Procedural work that involves generating and manipulating digital assets.
Situations where specialist knowledge is required to search or manipulate systems.
Be on the lookout for any of the above in your discussions and follow up with semi-open questions designed to explore the scope and impact of any opportunities. To give you a flavour, semi-open questions could look like any of the following:
“Can you give me some examples of the ad-hoc reporting requests you said were difficult to fulfil?”
“Can you walk me through that daily reconciliation process in detail?”
“How were you able to help the customer to solve that problem, eventually?”
“What information resources did you use to take that decision?” (“What else would you have liked to have had access to?”)
Save your closed questions for confirming specifics or checking your understanding before moving on to the next open question.
tips for good interviews
If you start discussing a challenge or issue that your interviewee is facing, try to avoid dreaming up potential solutions; instead focus on understanding the impact of the problem and what benefits solving it would generate.
The most valuable information is that which is provided spontaneously, without the interviewee being led. For example, never ask questions of the form “Would you like a widget that does X?” – instead ask a question that gives your interviewee the opportunity to tell you which problems they would like to have solved. Quantify the benefits by asking what solving said problem would enable them to do. Really dig into the impact and make sure that you understand it fully.
It’s good practice to complete your questions 5-7 minutes before the time is up. This gives you the opportunity to ask “Is there anything else you think we should have discussed today?” Many of the best ideas come from your interviewees; if you don’t give them the space to speak, you’ll miss out!
Don’t forget, at the end, to ask for permission to follow up with the interviewee should you have further questions at a later stage of the process. We often schedule second interviews during the Testing & Refinement phase when we are assessing our opportunity shortlist.
Try to capture the precise idioms and phrases used by your interviewees in your notes. When you present your findings in the final AI Strategy, couching it in the language of your interviewees demonstrates that you have listened to them carefully and understood their situation.
Assimilating the information
After each interview, flesh out your notes. As you progress, you’ll want to be compiling a “long-list” of potential use cases, constraints, questions for other interviewees and topics for further research. Highlight the points that crop up frequently and plan questions for other interviewees to explore them more deeply.
By the end of the interview process you should have enough knowledge to be able to sketch out a “long-list” of ideas for potential uses of GenAI with the following details:
What is the challenge, inefficiency or opportunity?
For a challenge or inefficiency: who is affected and what is the impact? What would solving it mean?
For an opportunity: what benefits could accrue if it was seized?
What data and systems are involved? What data are missing or problematic?
What blockers - technical, cultural legal or commercial - could prevent this idea from being developed?
What role could GenAI play in this?
At this stage, the long-list can be rough and the content can be incomplete. That’s fine because you’re going to develop the ideas in the next phase: “Research & Ideation” - which is going to be the subject of our next blog in this series.
For now, we hope this short guide to stakeholder interviews gives you enough information to try them out for yourselves. We wish you luck!