I have a strong opinion that the best use of AI is not in individual productivity enhancement - writing emails faster, or generating endless slop. Unexpectedly, I tried writing this with Claude and failed. It was generic and didn’t really hold my voice - I used AI to review, though - and in another article I tried again with Glean and the result was much better. Individual experiments in productivity don’t always scale, and worse, if you fail, it reduces your likelihood of using the tool next time. Adopting any new tool requires effort. We have to make it easy and more certain for someone under pressure that they will get a better result; it’s not unlike delegating work to an inexperienced colleague. I am convinced that if we want to get the best out of this amazing technology, we have to redesign our workflow to use it systematically and think about how teams and AI can work effectively together.
We are all accustomed, and somewhat immune, to friction in our day-to-day working processes. Take two examples.
1. The sales process.-
- An Account Executive (AE) preps for a meeting - they have to pull a report to get the history of the client, they investigate LinkedIn to see the contact’s previous history, and they look at news sources to see what’s happening. Then they go to the meeting, take notes, and update their CRM to make it accurate. It’s a disjointed, manual process.
- Applying AI on an individual basis to fix it is make-work - everyone tries to build the same thing; some succeed, some fail. You want to build an integrated, uniform workflow that uses AI to prep the meeting report and deliver it to the AE via whatever channel they prefer. The AE should be able to record the meeting, then an agent can pick up the transcript, action it, and update the CRM, perhaps also coach the AE on what could have been done better in the meeting. Design and build it once, roll it out to 200 AEs, then watch the employee satisfaction wins, savings, and customer experience improvements stack up. It’s worth being clear about where the risk sits: there’s a hallucination risk in the research against news sources, but not against your own data or in the transcription.
-
- A business analyst runs a workshop, writes up the output, creates user stories, reviews them and gets them signed off with the business. A developer picks them up, asks questions, and starts work on building. A tester picks up the same stories. The Business Analyst (BA) writes User Acceptance Testing (UAT) scripts, and eventually the business runs the tests. Every stage is lossy and slows down the overall end-to-end process.
- The first thing to recognize is this is a team job; it’s unusual for it to be just one BA, or one tester. Underpinning the software development lifecycle (SDLC) is usually a set of systems: Jira, process repositories, git, and so on. Each actor can individually improve using AI to review stories or write code, but the real improvement comes when you consider the whole end-to-end. As with the sales example, you want to build one integrated workflow that removes the drudgery and friction and reduces the number of different system interactions required. A BA can run a workshop and record it; AI can help turn the output into draft stories for the human to review and add insight to, do a preliminary review, and generate a lot of the downstream artifacts. Again, this has to be orchestrated and uniform, otherwise we end up building the same thing ten times.
What AI can’t do is replace the human negotiation, trust, collaboration, and re-imagining of how things might work. Our innately human ability to argue, persuade, and compromise becomes more critical as we automate away the things that soak up time and prevent us from achieving our full potential.
If you are facing a transformation that is moving too slowly, creating too much friction or losing momentum between strategy and delivery, get in touch to explore how CAM.ai can create a more connected flow for your organization.
