As we start 2026, artificial intelligence (AI) is no longer a curiosity sitting on the edge of our business landscape. It is already embedded in everyday work, either formally or informally. People are using it to draft papers, summarise meetings, turn rough notes into executive-ready updates and accelerate analysis that used to take hours or days. The adoption curve is real and it has been steep. Stanford’s 2025 AI Index reports that 78% of organisations used AI in 2024, up from 55% the year before. McKinsey’s global survey similarly reports 65% of respondents say their organisations are regularly using generative artificial intelligence (GenAI) in at least one business function, almost double the previous survey.
Project Management Offices (PMOs) sit at the intersection of delivery, governance, reporting and decision making. That puts us in a position to see two truths at the same time:
- AI use cases in the PMO are plentiful and genuinely valuable
- Scaling AI beyond individual productivity is difficult because portfolio data quality is inconsistent and governance is lagging
It is very easy to use AI tools in a project management context. Last year we launched an AI for PMs course that provides an abundance of use cases across all phases of the project delivery lifecycle. Using AI in a way that adds value is not the hard part. The hard part is scaling AI usage across the business in a way that is repeatable, decision-grade and safe. The difference is not the cleverness or availability of the tools, or even the maturity of the team. The difference is the quality of the data and the robustness of governance around how AI tools are used.
Why the PMO is feeling the impact first
Most project work now runs through digital platforms: work management tools, collaboration suites, knowledge bases, ticketing systems and reporting layers. AI is increasingly delivered inside those platforms, not as a separate specialist capability. Atlassian, for example, has embedded AI features in Confluence and a broader AI capability that connects knowledge across Confluence and Jira, supporting search and content creation within existing workflows.
This matters for PMOs because we already rely on those tools, and many other Project Portfolio Management (PPM) systems, to run the portfolio rhythm. Once AI is built into the tools, it becomes frictionless for individuals to use. A project manager can generate a first draft status update in minutes. A PMO analyst can summarise a set of project updates into themes. A coordinator can turn meeting transcripts into actions and decisions. It feels like an immediate productivity win. It is easy to switch this capability on and it is already happening in most large organisations.
While we accept the ‘AI-ification’ of our work platforms mostly without question, the tools generating the most buzz are large language models such as Microsoft Copilot and ChatGPT. These tools are making people far more efficient at the work that sits outside core platforms. Tasks like creating slide decks, which used to take days to craft, can now be done in minutes or hours. The analysis of large bodies of data, often spread across different formats, used to take days or weeks. Now it takes minutes or hours.
The catch is that general-purpose GenAI will do a wide range of tasks and it is often left to the individual to judge whether the input data is good enough to rely on. That is fine for meeting minutes, emails, or drafting a document. It is much harder for enterprise-level data where quality, completeness and definitions vary across teams.
Our challenge as PMOs is that we are not judged on whether an individual can draft something quickly. We are judged on consistency, quality and trust. When AI becomes part of delivery and governance processes, the PMO has to answer harder questions:
- Are the insights consistent across projects, programs and portfolios?
- Can leaders rely on what they are reading?
- Are we handling data appropriately?
- Are we clear about what AI is doing, what humans are doing, and who is accountable?
Systems and tools are everywhere. Capability is still rare.
From personal productivity to PMO capability
GenAI is spreading because it speeds up individual work such as drafting, summarising, ideation and quick analysis. It is widely in use, even where organisations have not formally sanctioned a specific tool. That first wave of adoption has been personal. Individuals use the tools to write faster, think faster, and tidy up the admin that slows delivery down.
The second wave is where it gets more interesting and often much harder. Individual productivity is not the same as PMO capability. Capability is when the PMO can apply AI consistently and safely to improve portfolio outcomes and reduce risk, in a way that stands up to scrutiny.
To get there, four foundations matter:
- Use cases: a small set of prioritised AI use cases linked to portfolio outcomes
- Data: decision-grade portfolio data with shared definitions and disciplined maintenance
- Governance: pragmatic guardrails that enable AI adoption without creating bureaucracy
- Operating model: clear roles, skills, quality assurance and routines for continuous improvement
Most organisations are doing well at identifying use cases and building skills. For the most part though, they are still playing catch up on data and governance. That is why enthusiasm and individual activity is rising quickly, while scaled and measurable impact still lags.
Many organisations are stuck in a pilot trap. Lots is happening, but the value does not compound because everyone is doing something slightly different and the foundations are not in place. If you want AI to scale in a PMO context, it needs to be designed like any other change: clear outcomes, good planning, standard ways of working, quality controls and governance.
The data reality: AI amplifies what is already true
AI tools are fantastic at turning text into cleaner text. They are far less effective at fixing weak data discipline and inconsistent portfolio controls. This is why data quality is the ceiling on PMO-scale AI.
In a PMO, poor quality data creates rework, reporting cycles that drag on, constant reconciliation, low confidence in dashboards, and slow decision making. Ultimately, it creates a loss of credibility with the leadership team.
Poor quality data often shows up as:
- schedules that are not updated consistently
- milestone definitions that vary between teams
- benefits described as narrative with no baseline or measurement discipline
- RAID logs (Risks, Assumptions, Issues and Dependencies) that are incomplete, or treated as compliance artefacts
- dependencies tracked in emails or corridor conversations, not the system of record
- resource data that is aspirational rather than actual
Individual use cases often succeed even when this data is messy because the output is still a draft. A human can correct it and move on. Data insights at scale work differently. Once you ask AI to produce portfolio-level information, compare programs, predict delivery risk, or identify systemic issues, data quality becomes non-negotiable. As AI is increasingly embedded into automated workflows, there may be less opportunity for a human to catch errors before they travel.
A simple test is this: would you be comfortable making a funding or reprioritisation decision based on the data currently sitting in your PPM tool? If the answer is no, then AI will not magically make it yes.
What “decision-grade data” means in a PMO context
Decision-grade does not mean perfect. It means:
- consistent definitions across the portfolio (baseline, forecast, progress, benefits, risk ratings)
- timeliness that supports the governance rhythm (weekly or fortnightly updates that are real)
- traceability so reporting can be linked back to underlying evidence
- comparability so leaders can see patterns and trade-offs without caveats
- ownership with clear stewardship, not shared accountability by default
When those conditions exist, AI becomes a force multiplier. When they do not, AI becomes an acceleration layer for inconsistency.
Governance is lagging because AI is moving faster than policy
The second constraint is governance. This is the part many PMOs try to avoid because it sounds like red tape. It does not need to be. Governance should be the mechanism that allows AI to scale through the organisation with confidence.
The external environment is tightening too. The European Union Artificial Intelligence Act is expected to apply in stages, with a general date of application referenced as 2 August 2026 for many provisions. Even for Australian organisations, these trends influence supplier behaviour, client expectations, and what ‘good’ looks like in risk management.
Two practical anchors are worth reviewing for organisations scaling AI practices:
- ISO/IEC 42001, which describes an Artificial Intelligence Management System and how to sustain governance over time
- NIST AI Risk Management Framework, including the Generative AI profile (NIST AI 600-1), which provides a pragmatic risk lens for GenAI
Australia’s AI Ethics Principles also remain a useful local reference point, particularly around transparency, privacy protection, fairness and accountability.
The message across these frameworks is consistent: the risks are not hypothetical and they are not in the future. They are here now.
What governance needs to cover for PMO-scale AI
Kept practical, governance becomes a set of guardrails that helps everyone move faster with less risk. In a PMO context, six areas matter most:
- Tool approval and deployment model
Which tools are approved, where they run, who owns them, and how they are configured. - Data handling and confidentiality
What data is allowed to be entered into a tool, what is prohibited, and what classifications apply. - Human review and accountability
AI can draft and suggest. Humans remain accountable for what goes to decision makers, clients and regulators. - Traceability and evidence
If AI produces a summary or insight, what sources does it rely on, and how do we show our working? - Quality assurance and monitoring
Are outputs sampled, are errors tracked, and is there a feedback loop that improves guidance and practice? - Decision rights
Where are the boundaries? AI can flag patterns. It should not make automated prioritisation decisions without explicit design and oversight.
The goal is not to constrain AI. It is to make it dependable.
A sensible PMO approach: standardise the right use cases
If you want to scale AI, treat it like any other change. Start narrow, prove value, then expand. A practical way to structure this is a three-tier use case stack.
Tier 1: safe, high-volume productivity
Low risk use cases where AI saves time without introducing decision risk, provided outputs are reviewed.
- meeting notes to actions and decisions
- first draft status reports, sponsor updates and project communications
- rewriting and summarising for different audiences
- converting long documents into short briefs
- generating templates and checklists based on PMO standards
This tier is also where you build good habits: prompt discipline, appropriate data handling, and consistent review.
Tier 2: portfolio consistency and comparability
This is where the PMO starts to see portfolio value. It needs stronger standards and improved data quality.
- consistent narrative across projects, including risk language and benefits language
- cross-program thematic summaries of risks, issues and dependencies
- standardised assurance summaries and action tracking
- structured extraction from unstructured updates into portfolio fields
When organisations get to this point, portfolio level discussions shift. Leaders stop arguing about whether the data is credible and start focusing on decisions and trade-offs.
Tier 3: decision support and predictive insight
Tier 3 requires the strongest governance and the best data.
- early warning indicators based on schedule, risk and delivery signals
- scenario modelling support for prioritisation and capacity decisions
- benefits risk prediction based on assumptions and delivery progress
- systemic risk detection across the portfolio
If Tier 3 outcomes are the goal, the foundations must be in place. Confidence in the tools and governance must be built from the ground up.
The PMO roadmap: practical steps for the next 90 days and next 12 months
Most PMOs do not need a grand AI strategy document. They need a simple operating plan.
Next 90 days: build momentum with control
- Choose five enterprise use cases
Pick high-frequency, visible use cases tied to pain points such as reporting cycle time, narrative quality, meeting load, and assurance throughput. - Create a short set of AI working rules
One or two pages is enough. Include approved tools, data handling rules, and minimum review requirements before anything is published. - Draw a clear “human review is mandatory” line
Anything that influences a governance decision sits above that line. - Identify critical portfolio data elements
Pick ten fields that drive confidence and enforce consistent definitions. Lift data quality without boiling the ocean. - Set up a light quality check
Sample outputs weekly, track errors and near misses, adjust guidance, and share examples.
Next 12 months: turn AI into a managed capability
- Standardise portfolio language and templates
AI performs best with consistent inputs. Templates, risk language, benefits language, and dependency definitions matter as much as any prompt library. - Improve the system of record
Reduce reliance on unstructured reporting in spreadsheets. Integrate delivery and reporting tools so the data AI consumes is current and comparable. - Build AI governance into existing PMO forums
Do not create parallel governance. Use existing steering and assurance forums to oversee use cases, risks and controls. - Develop capability across the delivery community
Train project managers, PMO analysts and sponsors in safe use, validation habits, and what “good” looks like. Use change management practices, like any other change. - Align to recognised frameworks where appropriate
You do not need certification to benefit from ISO/IEC 42001 or the NIST GenAI profile. Use them as checklists for governance and risk management.
What good looks like: measures that matter
If you want executives to back PMO-scale AI, measure outcomes beyond ‘people like it’. A practical scorecard includes:
- cycle time for weekly reporting and steering pack preparation
- quality of status narratives, measured through reduced rework and fewer clarification loops
- data quality trends for critical portfolio fields, measured through consistency and completeness
- decision confidence measured through sponsor feedback and fewer contested portfolio discussions
- risk outcomes measured through earlier detection and fewer late surprises
- evidence of improved productivity across the PMO and delivery community
Done well, and done consistently, AI-enabled tools can be a game changer for PMOs and delivery teams. Low-value, high-effort tasks reduce. Time and energy shifts to the work that actually moves outcomes.
Closing: capability is deliberate
AI adoption is accelerating across the economy and is already embedded in the platforms project teams use every day. For PMOs, the value is real and the use cases are plentiful. The constraints are also clear: scaling AI beyond individual productivity is difficult because portfolio data quality is inconsistent and governance is lagging.
If we treat AI as a personal productivity layer, we will get sporadic wins and inconsistent outcomes. If we treat AI as a portfolio capability, we do what PMOs do best: standardise what matters, lift quality, reduce risk and help leaders make better decisions.
Systems and tools are everywhere. Capability is still rare. In 2026, that is a huge opportunity for the PMO and project controls community.
How we can help
If you are serious about moving from ad hoc AI use to real PMO capability, we can help you do it safely and pragmatically. We work with PMOs to identify the highest value use cases, set fit-for-purpose governance, lift data quality where it matters most and embed repeatable ways of working so the benefits stick. That can include a short assessment of your current maturity, a prioritised roadmap across the three tiers, practical guardrails for tool use and data handling and hands-on enablement for your PMO team and project community. If you would like a clear plan you can action in the next 90 days, get in touch and we will help you turn AI enthusiasm into decision-grade portfolio performance.
References
- Stanford University Human-Centered Artificial Intelligence (HAI). AI Index Report 2025.
- McKinsey & Company. The State of AI in Early 2024 (Global Survey).
- Atlassian. Product documentation and announcements on AI features across Confluence and Jira.
- European Parliamentary Research Service. The Timeline of Implementation of the AI Act (June 2025).
- International Organization for Standardization (ISO) and International Electrotechnical Commission (IEC). ISO/IEC 42001:2023 Artificial Intelligence Management System.
- National Institute of Standards and Technology (NIST). AI Risk Management Framework and Generative AI Profile (NIST AI 600-1).
- Australian Government. Australia’s AI Ethics Principles.
