In 2025, AI can be an outstanding problem solving partner. It can also be a waste of time, and even a liability. To get the best out of this partnership means understanding where AI currently excels and where it flatters to deceive. It also means understanding what our role is as discerning human problem solvers. So that we’re smart users of this increasingly valuable but still limited tool.
Where is AI a Good, Bad & Ugly Partner?
Looking at where AI shines in the whole problem solving process, the picture is a mixed bag. It has some wonderful positives but some major gaps and potential trapdoors.
Let’s look deeper at those second and third stages. Here’s where we think AI needs intelligent and expert human beings for the process to shine.
AI as a Critical Thinking Partner
Here’s the critical thinking and research stages of the problem solving process in a bit more depth.
AI can help you with all of this, as long as you’re driving.
How Well AI Can Support Problem Solving
Without a clear critical thinker to tell AI what to do, your answer will have gaps. It will have no clear reasoning holding the conclusions together, and no obvious prioritisation into the most important things to understand and get right.
But AI can be the on hand hyper productive untrained clever clogs side kick. One that helps at the edges to give the critical thinker time and information to take their thinking so much deeper.
AI as a Research Partner
2025 research AI is a hyper productive graduate with no general domain expertise or quality control training who only looks at easily available sources. The graduate is also fully prepared to bullshit (aka “hallucinating”). If you managed such a real life graduate you would be very clear about your problem definition, precise with your delegation. You’d be diligent with your review and quality control of output. You’d also expect to take multiple iterations of this to get to an answer you’d bet your reputation on.
Here are the key steps in delegation. A clearly engineered set of prompts for an AI assistant can entail much more than this but this is a good list of things that must be in there.
Here are some important things to look for when reviewing the output from when you’ve delegated research to AI
- Source quality – has the Deep Research AI used sources you consider reputable or is it referencing some low quality market research outfit, opinion blog, or supplier with a dog in the fight? 2025 Deep Research AI has a big problem with this. It misses some excellent sources that are one click behind the website, unless you supplement with your own sources or do complementary direct research
- Source quantity and triangulation – is it relying on a single resource or methodology, or is it triangulating across multiple sources and methods? With 2025 AI, you need to be active with the triangulating. You can do this by yourself or as an additional instruction to the model
- Sense checking – do the numbers and trends it come up with make sense given your more general domain knowledge? Remember AI is taking what it’s given as facts. It builds a view from there, and there’s a lot of questionable “facts” out there
- Gaps – in the impressively laid out and extensive output, it’s sometimes hard to see the bones of what it’s saying. Taking the highest level outputs together, is that an exhaustive, or it least sufficient, list? Or has it missed something important?
- Prioritisation – AI can’t prioritise unless you tell it how to. Does it generate output focused on what’s really important and answer changing? What needs shedding as immaterial and what needs to be understood thoroughly?
- Key deep dives – Connected with prioritisation, AI can’t distinguish where it needs to go deep and where a headline answer will suffice. Where does AI need to go deeper in the next iteration, and what question do you need to ask to achieve this?
2025 AI as a Problem Solving Partner
With the right hand holding, like that smart, hyper productive graduate with no domain knowledge, 2025 AI can be a wonderful problem solving partner. Just make sure that you own the critical thinking, and that you keep the delegating, reviewing and quality control on a tight leash.
