The rise of the AI-powered PMO
…] Everything that we have electrified, we are now going to cognify.And I would suggest that the formula for the next 10. 000 startups be very very simple: take X – and add AI.” – Kevin Kelly, WIRED
Introduction
In February 2019, I applied to be a presenter for the PMI Summit organized by the Medellín, Colombia Chapter. I was selected, and that October, I presented my lecture on the future of the Project Management Office (PMO) to dozens of PMPs and project professionals.Alas, time flies! It has been more than five years since that event. The incredible progress of Artificial Intelligence, particularly Large Language Models (LLMs), is paving the way for the fulfillment of many of the “prophecies” I shared that pre-pandemic day in beautiful Medellín. It is time to refresh those ideas and tweak that vision based on current developments.
The lecture
The underlying principle of my 2019 speech was simple: Project Management and its related disciplines—PMOs in particular—were on an irreversible path toward automation through increasingly advanced computerized systems. I predicted four phases:
1. Integration & Automation: Using RPA and “dummy-proof” workflows.
2. Chatbots: Virtual team members acting as user-friendly “wikis.”
3. Machine-Learning Based Project Management: AI analyzing data via neural networks to discover hidden trends and suggest predictive courses of action.
4. Fully Autonomous Project Management: this is the complete automation of the project manager role, something foreseeable at least for simple projects with few stakeholders.
Let´s elaborate on those stages. Phase 1 was not truly AI-powered, but a mix of Robot Process Automation (“RPA”), dummy-proof workflow enablement, “Groupware“ massive adoption and other items. Phase 2 was about the enablement of chatbots within our communication tools, some through text, some using voice, acting as virtual team members and user-friendly equivalents to “wikis and knowledge repositories. Phase 3 is where things are supposed to get interesting. Chatbots become readily available even in wearables. AI steps in and analyzes the data through powerful neural-networks, discovering trends and relationships invisible to the human eye: conclusions are now predictive and the AI suggests courses of action to the PM. I also anticipated something I still linger for: actionable outcomes, updates and tools derived of project logs, history & lessons learned. Then Phase 4 is defined as to be understood as the equivalent of a Level 5 Automation grade as per the standards of the Society for Automotive Engineers (SAE) scale of self-driving. The speech also included mentions to the impact of drones in construction projects, the Volatility-Uncertainty-Complexity-Ambiguity nature of our moder world (VUCA) and concluded with the need of ever improved soft-skills from professionals in general and PMs in particular, and the importance of “defending” our human nature. That was more than five years ago.
Hits & Misses
Let´s begin with the latter. My analysis ran under the assumption that those phases were going to occur basically in a “step-by-step” fashion. That was a mistake. As William Gibson famously said, “The future is already here – It’s just not evenly distributed”. He was right: the phases mentioned are running simultaneously, or at least with strong overlaps. There are also different levels of progress across different geographies, sectors and industries. And then there is back & forth, a tremor in which progress is made but then a crisis, an error or backlash halts or even setbacks the process for a while. For example, at this point I thought that powerful & knowledgeable bots should be ever present across organizations of all sizes & types, and that AI-powered agents should be available within internal project spaces, automating a bunch of the tedious, repetitive work PMOs and PMs execute. This is only partially true: it depends on the industry, country, sector, organization size and type.
Now, even if I mentioned the ever more “VUCA” nature of our reality, I could have never anticipated the degree that this has reached. Wars, international rule order disruption, immigration, energy scarcity… you name it. The “VUCA” term was coined by the Pentagon in the mid-80s and became popular during the first decade of the 2000 century, but the “mess” went off the scale with the pandemic and subsequent major events (the Ukraine war, tariff changes, chips war, Iran war, etc.). These events add chaos layers over each other. The ultimate consequence may be global stagflation, even a new Great Depression. We will see.
I also missed the non-linearity of this process. If we graphed our progress, it wouldn’t be a straight line; it would be a “hockey stick.” We are just reaching the curve where progress becomes exponential. It took time and effort to get to that point, but here we are, empowered all across the place through AI. However, my prediction on an equivalent of “augmented reality” for PMs and PMOs is not yet around us.
Let´s analyze now the hits. I had a big-hit with Groupware. MS Teams, Slack and other similar software is now omnipresent even for small organizations, and meeting through a screen is a daily thing (this was turbo-charged by the pandemic). Hybrid project management has become usual, and LLMs regularly check contracts, emails and documentation in general. Drones are now ubiquitous and regularly used for a variety of needs: civil engineering, topography, agriculture, military, security, etc. But indeed my “home-run” was the evolution of the value added by PMs & staffers in general. AI & software is each day more capable and is taking over tedious, repetitive tasks and in general work that does not requires critical thinking. This means that soft-skills and an elevated level of judgement is ever more important.
What comes ahead for PMs and PMOs
Its 2026. As a whole, my predictions were off by five years, thus an automated PMO should be available around 2030. What do I mean by “automated”? Well, PMOs come in different flavors and sizes. I think that in five years the amount of collected data, the recursive nature of algorithms´ improvement and organizational maturity will enable the first PMOs that run more than fifty percent of their processes using agents and LLMs. Some of the best candidates for automation are:·
- Induction & Training
- Gates´ reviews & approvals workstream (not Bill, but the project progress thresholds :-)·
- Documentation creation & versioning·
- Predictions with probabilities and ranges: Cost, Duration, Risks
- 24×7 AI powered support agents & bots providing answers, tools and help for the staff to better manage projects; programs and changes. Examples: staffing details, task info, RAID analysis, blueprints, timelines & schedules, costs, scope, performance, monitoring & controlling, quality, etc.
However, there are aspects of PMOs that cannot be fully delegated to an AI. Agents are now here and some decisions can be delegated to them: it’s a matter of defining which decisions those should be and calibrating them to the organizational risk appetite, industry & governance model. However, in my perspective, there is a limit to what should be delegated to AI. Decisions that imply medium & high impact should be overseen, if not driven or entirely made by human beings. This does not mean removing the outputs from the AI analysis off the decision process, but to include those as inputs for a more comprehensive evaluation that includes human judgement & instinct: think of it as a “cyborg” that merges the deep trend & relationships analysis of AI with human common-sense, big-picture perspective and sensibility. It’s a winning combination.
I also think that these changes imply challenges to organizations, particularly from the governance perspective. Who would be accountable for AI powered decisions, particularly if they go wrong? What type of decisions should those be? What are the limits, controls, thresholds, derailers and alarms associated to those decisions? Who should audit those decisions, and with what frequency? What AI models & agents are to be used – internal or external? Where should the data be located? What security aspects should be considered when operating under such a model? What are the legal implications? What information should be communicated to shareholders & other stakeholders? What would be the correct approach for the “moments of truth” – the moments to make decisions? Eg, should the AI agent have a single vote such as the persons in the board? None? Several? How and when to vote? Ultimately, we may need to ask ourselves about the very purpose of PMOs, meetings and decision sessions Nothing is off the table
Conclusion
In short, it looks like my crystal-ball was showing me somewhat fuzzy images, but lately the images are becoming reality. PMOs have a brighter future in which lessons-learned are actionable, predictability is built into the system and information is readily available in user friendly interfaces. At the same time, these changes disrupt organizational governance and accountability, thus deep thinking and preparation is required. Ultimately, this evolution is inevitable and required to navigate an ever more volatile, uncertain, complex and ambiguous world. Tell me, what are your thoughts on this?
Best human vibes,
Fernando
Decisions, Procedures & HIPPOs
If there´s a moment in the organizational day-by-day that majestically embodies the term “Governance” that is the moment of making an important joint decision. What amazes me day in, day out is the absence of an “architected” approach to those moments of truth. In my experience, regardless the size of the organization, its industry and maturity level decisions are quite generally taken using a “primitive” procedure that exists just due to momentum and lack of critical thinking. In the coming lines I intend to raise this situation to the reader´s awareness and provide some food for thought for you all.
Let´s start with an example. July, CIO, is presiding the monthly IT Security SteerCo for ABC Enterprises. As per the recent increase of cybersecurity attacks, the committee must make a decision. They need to increase their IT security level. Thus, they need to choose between several IT Security suites and providers. The meeting begins somewhat late and there are solely 30min booked. To make things worse, conversation digresses and when the topic comes up, July takes command of the call and states her preference. Tommy, CSO, has concerns about the suggested solution, particularly with the 3rd party implementer. Mark, IT Ops Director, too, but more from the solution itself perspective. He has heard negative feedback from peers in other organizations. Luke, CTO, has no particular position. Ditto for Emiley, PMO Lead. The clock relentlessly spins its arms and after some discussions, the “Five minutes left in meeting” alarm pops-up in everyone screens. July takes again command of the meeting and states: “Okay, let´s get to a decision – every day that passes we are at risk, we must not postpone this anymore. I also have to jump to another call. I vote for the mentioned software, and I can have our friends from SuperDeploy next week on-site to define the implementation approach in order to get a formal proposal. So what is your final take, Tommy?” Tommy feels the pressure, he´s put on the spot. So he concedes. Mark makes a couple final rebukes but alas, “if July and Tommy agree, well, me too. So, yes”. The rest of the team robotically say “yes” and that´s it. The team adjourns the phone bridge. And we all say, “Geez, so, what just happened here?”
If you were paying attention, I guess you came to the conclusion that the aforementioned meeting was engineered to fail. Or more precisely, it lacked engineering. It was an ad-hoc, impromptu improvisation with no script, no guide, no agenda and no method, particularly for the moment of truth (the voting exercise). This is my core point: smart, logical, fact-based decisions are not taken like this. The scenario was perhaps exaggerated (lack of punctuality, lack of focus, short timeframe, etc.) but I think that we all have seen stuff like this in our careers. The voting exercise is the summit and culmination of it all. Once July makes her preference utterly clear and public (and that is the Highest Paid Personal Position or HIPPO) there is nothing left to be said. Her CIO role and commanding style pushes the rest of the attendees toward her preference. And the decision is taken.
A wise man makes his own decisions. An ignorant man follows public opinion
Chinese Proverb
So what´s to be done? I don´t have a one-size-fits-all answer, but First thing is to be aware of this issue which seems to affect us like a chronic disease to which we have become anesthetized. Secondly, I believe that this is the type of meeting (and particularly in a WFH / Remote culture) is the one that demands a more strict business approach: punctuality, pre-defined agenda, pre-defined time-slots, pre-defined priorities, pre-defined roles (eg, note-taker, chairman, etc.). And Third, I say that voting must not be taken lightly. The HIPPO(s) must be left at the end, for obvious reasons. If possible, the possibility of simultaneous voting tools should be considered, and perhaps in some cases, even private voting. Disclaimers should be warranted. Even second debates for final endorsement, maybe, in business critical matters. It´s a matter of creativity and tweaking an adequate solution for each org.
At the end, I believe the point is clear now: “Management is doing things right”, said Drucker. Let´s make sure that the second part of his quote – “Leadership is doing the right things” works out during voting exercises.
Fernando
