A consulting and training package that redesigns high-stakes workplace workflows so artificial intelligence supports human judgment instead of replacing it.
Added Jul 26, 2026
Organizations are encouraging artificial intelligence adoption without clear rules for which tasks require independent reasoning, human review, or restricted tool use. Employees can generate polished business cases, analysis, and written work without understanding or testing the underlying claims, creating mediocre output and hidden decision risk.
Deliver a fixed-scope audit of one recurring workflow, such as preparing business cases, policy recommendations, or client reports. Map where artificial intelligence is appropriate, establish required human reasoning and verification checkpoints, create task-specific usage guidance, and train the team using examples from its own work.
Rapid adoption has made it easy to produce plausible work before organizations have developed mature usage practices. Concern is shifting from whether employees use artificial intelligence to whether they can evaluate its output and preserve accountable human judgment.
Showing 1-11 of 11 signals
So actually having a mature approach to AI would, I'm answering my own question, help prevent some of these things from happening. So I can totally see and understand why having a kind of framework around suitability for different models to do different things for different people would make a huge amount of sense. And I think just getting people to be open about using these things and using it in the right way, not, not, I came up against this wonderful phrase by a lady called Sol Rashidi, who calls herself the first chief AI officer. And she's got wonderful content. She talks about AI atrophy and saying that you can just treat this stuff to outsource your thinking. It's so easy now to generate anything. Write me a business case. It'll be 30 pages long. It'll take two minutes. You don't do any thinking for that. It's going to be very mediocre. And all you're doing is you're just pushing the work somewhere else. So I think if people are worried about people adopting AI in those types of ways, they're worried about the mediocre output, then it's a risk. But if you go down the route of, well, let's help people use the right tools for the right jobs and use it for brainstorming. Don't outsource your thinking. Use it to critique. Use it as a collaboration partner, as a mentor, these types of things. Then I think people will instinctively know through use, oh, OK, maybe I shouldn't upload this information. And I think you just sort of remove that concern through education and engagement. I think this is why that one of the most future-proof skills is critical thinking. It's the ability to question everything. Even the feedback that you get back in mouse might sound super intelligent, like it's produced this 30-page business plan. But you still have to have the nails to be able to say, well, yeah, I kind of see where you're getting at, but that's not the market that we're aiming for and because blah, blah, blah. So you still have to reasonably, I don't think you can fully delegate anything and suppose that it's the master of everything. Yeah. Is that a fair comment? I mean, yeah, I would agree. I guess that people talk about it, AI as a great sort of multiplier tool. But what's the multiplying? If you don't put any effort in, you know, you just put in a one-line prompt, right in your business case about X, you're going to get very average outputs.
What are the top three things you should never use AI for? I would say any skill you want to preserve in your head, you should probably not use AI for it. So I use AI for editing right now. You very quickly end up on a slippery slope. So if I create a rough draft as I did with the self-help trap, for instance, I would then take that, feed it into these models and give them a personality. You are an editor from the New Yorker. This is your name. Maybe it's a famous editor or the person who worked with Robert Caro, whatever it might be. I mean, that's, again, not to compare myself to those people, but I want a good editor. Give me feedback on this rough draft.
My copy editor, GEM doesn't say anything about rewriting it or suggesting ideas. It just says, proofread this for grammar, spelling, and clarity. Do not rewrite it. Do not rewrite anything. Just tell me the changes to make, because I am a better writer than you. So we don't want AI influencing our thoughts, our writing. Why would you want to outsource the thing that makes you uniquely you? Another way to think about it is like when you shallowly think of something and then outsource the rest to AI, you're not making your own ideas come to life. You're making other people's ideas come to life. So if you use AI sparingly, you won't get left behind.
One00:38:15caution in the generative AI space is, I would say, maybe outsource your writing, but don't outsource your thinking. Writing for me is thinking, so, you know, I cannot outsource my writing as well. And how would you put this in perspective is very important. Oftentimes00:38:30and the second one is, you know, make sure your writing is actually read, and you get feedback. Someone should not be taking your writing and then putting in a summary and then consuming the summary of it. I think you need to write in such a way that people need to consume it raw. And this is important in the age of agentic AI, I would say. At some point, it would become pretty normal that people would stop sharing just the data, but add a commentary, add a summary, and that becomes a new normal.
If I was working with a, junior strategist and we were sparring back and forth, it's still at the end of the day, it's my responsibility to vet the information that I'm getting to put it together in a way that you can then present to the client. And the client is depending not on that junior strategist expertise. They're depending on mine. Yeah, totally. And, I feel like it's, it's the same with AI. Now I do think there are ethical concerns with just copying and pasting directly from, from the platform straight into a document and sending that off for approval. Like those are not
+8 more signals