How have you calculated ROI of AI solutions (including agents) that you've rolled out at your firm? Are there specific KPI's that you've focused on and how have you measured (and validated) the value?
Sort by:
no title2 days ago
I can only echo this. The only thing we did at one point in time was e.g. comparing things that are clearly measurable. So we enrolled entire teams on github Copilot and compared their Software Delivery KPIs overall some months before and some months after. There are clear indicators that teams are more productive (velocity increased from as low as 5% to as high as 50%) as well as code quality slightly decreasing. We also include factors like team happiness into that equation. Long story short: what we were looking for is a reason to justify the expense and that has been found. Cause even if a team only gets a couple of percent more productive there is a strong argument to cover the cost associated.
This is a multi-facted beast. Gartner did some interesting research on it, splitting application of (Gen)AI into stuff that makes current processes run more efficiently (think e.g. MS Copilot, ChatGPT, etc.), stuff that adds new capabilities and changes/extends current business processes to generate more value, and stuff that is transformational and opens up entirely new processes, value streams, business models.
The argument is that financial ROI generally makes sense for the latter two, but less so for the fist category. There, it is mostly employee satisfaction / NPS. So think about that as return-on-employee rather than return-on-investment. The latter two categories are business cases like introducing any other new capability.