You were saying that AI is not a gaffer, but a mate that can help

Indeed - the phrase Microsoft are using for AI in software development is "co-pilot".

At the moment it feels like an enthusiastic, well presented, but ultimately a pretty dim apprentice. It's good a finding things other people have done and can repeat them parrot fashion, even pretty good a dressing them up to look very plausible indeed. What it's still seems to be very poor at is actually understanding if what it's found actually useful for the problem in hand.

What I've seen so far is perhaps 80% accurate and helpful - but the remaining 20% is not only dross, but so well presented that it takes an inordinate amount of human time to filter it out and find a useful substitute, meaning it was probably quicker just to do it yourself in the first place.

I suspect normally we rely on many subtle human clues when talking to people to gauge the value of what they're telling us and so can intuitively home in on useful information fairly quickly (e.g. if you ask a stranger for directions, and when answering if they appear drunk or say left while their hand makes a gesture to the right, you're probably not going to trust what they say; or when reading text if the spelling and grammar seems all over the place it might raise a worry, or if it contains the odd snippet that just doesn't add up, or you know it's from someone who has proved unreliable in the past. Stuff generated by AI seems to be anonymised and stripped of all such clues.

   - Andy.

Regardless of what we think, AI is inevitable and unstoppable, and that it will change the world in ways that we cannot control. Although I am bit pessimistic and resigned about the future of AI.

Regardless of what we think, AI is inevitable and unstoppable, and that it will change the world in ways that we cannot control. Although I am bit pessimistic and resigned about the future of AI.

I agree with you.  AI is gaining traction and will definitely affect the way we all work in Project Management.  Contrary to the PMBOK, no project is truly unique.  Whilst context can be unique a project is a collection of subsystems.  For instance, if looking at a building one sees a collection of various systems and subsystems which are always generic in nature.  Take a HVAC system.  It has a motor, filter, pump/fan, dryer, ducting etc.  On a generic level each should have a generic DFMEA etc which may not be bespoke to the project in hand but is a valuable source of risk management.  From such a dataset a model is constructed (i.e digital twin) from which with variable design, operational, and MRO parameters are input...thus parallel engineering though concurrent simulation results.  The use of AI here greatly speeds up the identification of the optimum design and relevant safety case.  In addition, AI also greatly speeds up the Critical Path once the option WBS is identified.

Personally I see not only AI driving greater efficiency for the Project Manager, but the PM will also have to greatly raise their game.  This will no doubt be reflected in any competency framework for Project Managers in the future.

I can certainly see the benefit in scenario modelling, given typical project managering of trying to balance the three strands of running late, and/or over spending, and/or failing to fully deliver, an aid to realistically modelling the "what if" cases would be good. However I can see the challenge of having valid data for the AI to work with - for occasional large infrastructure projects (we don't build 1000 HS2s a year) or occasional innovation projects it's an interesting question as to where the AI will get its source data from.

However, I agree with the comments above that I would still expect the final decision to be a human one, just because a lot of project management decisions are going to be the "wrong" answer for at least one stakeholder, and so I guess that we'd want a human to be able to be held accountable for that final decision. Even if it's based on AI data - but I'd hope that an effective AI system wouldn't give a single "correct" answer (which there never really is in PM) it would give a series of possibilities and resultant consequences to choose from.

I agree Andy.  AI certainly is not the 'be all and end all' in project management and never will it be.  Scanning technology is a great aid to build a 'digital twin' from which to simulate and can prove to offer significant advantages but as with all such architecture it must be verified and validated.  The reaction of the 'model' must therefore be aligned to the real world with suitable known calibration parameters.

You cite HS2 as a possible example for challenge.  Agree that AI could not be applied to the whole project but it certainly does have applications with bespoke elements of the project.  For example:

Tunnelling equipment optimisation and maintenance repair and overall (MRO) strategies.  This links well to the work being done in the Integrated Vehicle Health Management (IVHM) agenda and the work being done in prognositic health management (PHM),

The solutions offered by IVHM, PHM, and Through-life Engineering Services (TES) lend themselves well to the application of AI when looking at High Value, High Risk, complex projects.

Whilst AI offers a paradigm shift I would not advocate the use of 'closed loop' thinking or decision making within such architectures and models for obvious reasons.  Systems should always be 'open loop' with the human as the moderator.