AI, especially generative AI, is on everyone’s lips. What is it? How can it be used? How do we get real business value from it? With these questions and many more in mind, we sat down with our team and started our expedition towards the (fairly) unknown.
My team is responsible for the development of Gofore’s internal resourcing tool, Hohto. Hohto offers tools for skill management, allocation management, and CV printing, among others. It is central to Gofore’s business success, so our visibility into the different needs across our organisation is, for good reason, quite broad. Thanks to this visibility, we occasionally have the opportunity to support our colleagues in entirely new ways – like you’re about to see.
Where did we start?
The high-level goal was to gain experience in using AI for problem-solving and to identify potential opportunities where AI could improve user experience and create value. We approached our goal using a traditional 5 steps approach: workshop for knowledge and ideas, quick proof of concepts with decisions, design, test with a limited audience, and release.
We also defined a couple of hygiene factors, such as evaluating the proof of concept (POC) from ethical, data privacy, and EU AI Act perspectives. For potential wide releases, we also noted the importance of constant monitoring, adjusting, and updating after deployment, as well as collecting user feedback on a regular basis.
Use case: AI bot as a consultant’s partner in interviews
In the brainstorming phase, one of the questions identified was: how could we support our experts in customer interviews? Articulating one’s expertise in interview situations can be challenging: the stakes are high, time is limited, and others are watching. It can be a tough situation even for the most experienced professionals, as customers can approach interviews in various ways. Yet communicating one’s core skills, strengths, and areas for development is a crucial part of every consultant’s skillset.
Human-to-human sparring for individuals and teams is important, and we do a lot of it at Gofore. However, we started to wonder whether we could offer additional support from a completely new angle: AI. This led us to the idea of an AI interview sparring bot: a low-threshold, easily accessible tool available anytime and anywhere! The idea to test was if AI could be used to help experts prepare for upcoming customer interviews by creating elevator pitches. Alternatively, could AI spar with experts in other ways to improve their interview readiness?
What was done?
We already had experience with different large language models, so it was easy to select a suitable one for this use case. Our organisation has well-documented guidelines for preparing for customer interviews, so we used those as a starting point.
Initially, we created prompts based on our internal customer interview instructions. The results, however, were not sufficient. We then tried adding general, high-level project information as context for the prompts. Importantly, no data protection or actual customer information was ever compromised. With this adjustment, the results improved significantly.
In our third experiment, we tested whether a large language model could generate example answers to questions proposed by the user. This was done by providing the candidate’s existing sales pitch and general project information as context. The results of this experiment were also positive, encouraging us to continue developing the AI interview bot.
In this experiment, the AI bot was text-based. However, in the future, it could also support verbal conversations between the bot and the expert.
The results
Based on the outcomes, our hypothesis was validated. We successfully built a POC that uses the consultant’s profile information and basic project details to simulate interviews and spar with the consultant, helping them refine their customer interview skills. While face-to-face sparring remains the primary method for preparing for specific customer interviews, the bot could be a valuable tool for anyone looking to enhance their general professional skills, especially in articulating expertise.
However, it is important to remain mindful of AI’s potential shortcomings, such as hallucinations – instances where the AI generates incorrect or misleading results. Users must critically evaluate the tool’s outputs and view it as an additional layer of support to their own expertise.
What would we do differently or what would be a level up for the bot? Seeing how far we can go with this would definitely be interesting! Having more time for ideation and testing different hypotheses would be great, but that was not what this exercise was about.
As a level-up? We could have an avatar with a selection of voices and personalities for the user to choose from. How about an over-enthusiastic interviewer with the voice of our CEO? Or a never-satisfied, seen-it-all type of interviewer having your own voice as the audio experience? The confusion would be guaranteed, as would be your real-life survival skills in surprising situations after those sparring sessions!
You can also listen to this blog
This audio file has been created using AI technology that converts text into speech.
