Tuesday, October 10, 2023

PMO co-pilot

It seems that every PM framework and tool set now has a co-pilot app fired up by a generative AI engine of some kind. And, it's not just our domain: MS Windows, various cloud services, and others all now have the requisite co-pilot.

In a few words, generative AI has arrived big time in project management for many of the same reasons it's arrived elsewhere:
  • Speedy results, if there is enough training data. And not just speedy, but incredibly fast by most standards of creativity
  • Deep dive data driven, not just a surface look. Stuff you may not have known was there is, in fact, there.
  • Multi-variate processing; if you have enough richness in the data, then generative AI can consider huge numbers of parameters when synthesizing or optimizing a result.

All three of those are big-time influencers in the PM world. Speed (with accuracy) may monetize right to the bottom line. Not only from labor savings, but from opportunity capture and exploitation only available from acting fast.

The big query+
PMOs have had data warehouse capability (big data storage, and complex retrieval query engines) for many years (decades, really). But this is very different because the query engine is not some SQL-based application, but is really a neural-network inquistor that is manifestly more powerful in it's ability to consider scope (data + parameters + alternatives), and synthesize outcomes that nowhere existed in the data warehouse. (Obviously a strength and weakness: truly innovative, but how do you validate an outcome that nowhere else existed?)

And so the query power, if you want to think in data warehouse terms, brings on a big shift in division of labor; some jobs will be demoted or disappear and others will be created out of whole cloth .. to wit: prompt engineer.

Assisted Brainstorming
As a practical matter, a simple way to start using a co-pilot in the PMO is to use the AI engine (co-pilot) for assisted brainstorming. It's really really useful in that task. Give it prompt; evaluate the answer; and then prompt again to expand or explain. And on it goes. In just a few minutes you may have litterally pages of ideas to drive project ideas. 

All things considered
The big deal for project management is the amazing scope -- that is, the number of influencing parameters -- that can be cranked into schedule optimization, cost estimation, and risk management. 
  • For schedules, you've got schedule logic, but then parameters for risk elements, resources, immendiately past productivity (cost and schedule efficiencies) and environment (tools, support systems, virtual connectivity, even the weather). There could be literally multi-thousands of optimizations evaluated, given the broad scope of parameters.

  • Resource parameters could include various talent-indicator parameters, sensitivity to teams and direction, track record re problem solving, drive and initiative assessments, etc. In some ways, a "big brother" input to scheduling which your 'boss' already has in their mind's eye. (This begs the question: how does these parameters get regulated and validated from an HR perspective?)

  • Cost history can be "adjusted" to current parameters to make history more relevant to the instant situation.

  • Risk management can be made more calibrated and therefore more quantitative, putting less emphasis on qualitative opinions and hunches. All manner of risk tools which are parameter driven can be made more robust and more relevant faster.

  • For monte carlo simulations, and similar simulation and statistical considerations, like optimizing Bayes 'guesses', a generative AI tool can create infinitely more data points, data patterns, and trends which are then evaluated for their statistical properties with monte carlo techniques.

And then there's the paperwork
Every PMO is backed up by reports, with data entry coming from all points, and multiple readers who want not only executive summaries, decision trees, and red-green indicators, but also full-length explanations and backup.

At some point, it's a matter of a prompt for which then a product pops out!  All on schedule; all produced, validated, and distributed without intervention.  This may be a good thing to have.

Like this blog? You'll like my books also! Buy them at any online book retailer!