Imran AfzalAuthor| The Strategic PMOCary, NC, United States
I recently finished PMI's new The Standard for Artificial Intelligence in Portfolio, Program, and Project Management and found it to be a thoughtful framework for governance, risk management, ethics, stakeholder engagement, and responsible AI adoption.
One theme that stood out to me throughout the document was the emphasis on data quality, human oversight, accountability, and decision-making.
As I reflected on the standard, I found myself thinking about a related question.
Most organizations don't make decisions directly from raw data.
They make decisions from interpretations of data:
Dashboards
Reports
Metrics
Summaries
Recommendations
Executive briefings
Historically, those interpretations were created primarily by people.
Increasingly, AI is participating in that process.
AI can now generate meeting summaries, portfolio analyses, risk assessments, prioritization recommendations, executive updates, and decision-support artifacts.
Which raises an interesting question.
As AI becomes more involved in generating the information leaders consume, should organizations be paying as much attention to the quality of interpretations as they do to the quality of data?
In other words:
If data quality has traditionally been the foundation of good decision-making, how should organizations think about validating AI-generated interpretations before those interpretations influence decisions?
I'm curious how others are approaching this.
Do you see the next challenge as primarily a data quality problem, an AI governance problem, or something else entirely?
Saving Changes...
Sort By:
Luis BrancoCEO| Business Insight, Consultores de Gestão, LdªCarcavelos, Lisboa, Portugal
An excellent question.
I would argue that the next challenge is not primarily data quality or interpretation quality. It is decision quality.
Organizations rarely act on raw data. They act on interpretations of data, filtered through context, assumptions, incentives and human judgment.
AI introduces a new layer into that process. Even when the underlying data is accurate, an AI-generated interpretation can still be incomplete, misaligned with context or overly confident.
For that reason, I believe organizations will need to validate not only data quality, but also interpretation quality, contextual relevance and decision rationale.
The future challenge is not simply governing AI models. It is governing how intelligence is transformed into decisions.
In that sense, trustworthy AI requires trustworthy data, trustworthy interpretations and accountable human judgment.
...
1 reply by Imran Afzal
Jun 21, 2026 8:01 PM
Imran Afzal
...
Luis, I really like the distinction you're making between interpretation quality and decision quality.
The phrase "governing how intelligence is transformed into decisions" particularly resonates with me.
It also raises an interesting question: if AI increasingly participates in generating the intelligence, recommendations, and summaries that leaders consume, how do organizations determine whether a decision was influenced by poor data, a flawed interpretation, or simply human judgment?
That attribution challenge feels like it could become increasingly important as AI becomes more embedded in organizational decision-making.
Thanks for the thoughtful perspective.
Saving Changes...
Imran AfzalAuthor| The Strategic PMOCary, NC, United States
Jun 21, 2026 10:50 AM
Replying to Luis Branco
...
An excellent question.
I would argue that the next challenge is not primarily data quality or interpretation quality. It is decision quality.
Organizations rarely act on raw data. They act on interpretations of data, filtered through context, assumptions, incentives and human judgment.
AI introduces a new layer into that process. Even when the underlying data is accurate, an AI-generated interpretation can still be incomplete, misaligned with context or overly confident.
For that reason, I believe organizations will need to validate not only data quality, but also interpretation quality, contextual relevance and decision rationale.
The future challenge is not simply governing AI models. It is governing how intelligence is transformed into decisions.
In that sense, trustworthy AI requires trustworthy data, trustworthy interpretations and accountable human judgment.
Luis, I really like the distinction you're making between interpretation quality and decision quality.
The phrase "governing how intelligence is transformed into decisions" particularly resonates with me.
It also raises an interesting question: if AI increasingly participates in generating the intelligence, recommendations, and summaries that leaders consume, how do organizations determine whether a decision was influenced by poor data, a flawed interpretation, or simply human judgment?
That attribution challenge feels like it could become increasingly important as AI becomes more embedded in organizational decision-making.
Thanks for the thoughtful perspective. Saving Changes...
I think organizations should pay attention to the quality of AI interpretations, as flawless data can still lead to disastrous decisions if an AI misinterprets trends, hallucinates risks or applies biased framing. Because these outputs directly influence leadership actions, the challenge is having the right set of data and interpretation to make best decisions.
To make the best choices, decision-makers must have full visibility into how the insights were generated, explicit confirmation that all boundary conditions have been covered and proof that no underlying bias exists. Organizations must treat AI generated reports as unverified drafts, establishing clear parameters for what AI is permitted to analyze and requiring absolute traceability back to the source data. To validate these interpretations, companies should implement human in the loop, where domain experts audit the logic of AI generated summaries and executive briefings before they reach decision makers. There should be a strategy and process to review and compare historical scenarios to ensure consistent, unbiased and context-aware reasoning.
Data quality remains the foundation, but robust AI governance acts as the critical guardrail ensuring that interpretations are transparent, comprehensive and entirely free from bias so that leaders and decision makers can make the best decisions.
...
1 reply by Imran Afzal
Jun 22, 2026 5:13 AM
Imran Afzal
...
Srikana, I really like the idea of treating AI-generated interpretations as unverified drafts and requiring traceability back to the source data.
Your point about human-in-the-loop review is particularly important.
What I continue to wrestle with is scalability. As organizations increasingly use AI to generate summaries, analyses, recommendations, and executive briefings, there may come a point where humans can no longer validate every interpretation in detail.
At that stage, trust becomes a critical organizational capability. The question may shift from "Can we validate every interpretation?" to "How do we determine which interpretations require validation and which can be trusted?"
Thanks for the thoughtful perspective.
Saving Changes...
Imran AfzalAuthor| The Strategic PMOCary, NC, United States
Jun 22, 2026 12:20 AM
Replying to Srikana Ray
...
I think organizations should pay attention to the quality of AI interpretations, as flawless data can still lead to disastrous decisions if an AI misinterprets trends, hallucinates risks or applies biased framing. Because these outputs directly influence leadership actions, the challenge is having the right set of data and interpretation to make best decisions.
To make the best choices, decision-makers must have full visibility into how the insights were generated, explicit confirmation that all boundary conditions have been covered and proof that no underlying bias exists. Organizations must treat AI generated reports as unverified drafts, establishing clear parameters for what AI is permitted to analyze and requiring absolute traceability back to the source data. To validate these interpretations, companies should implement human in the loop, where domain experts audit the logic of AI generated summaries and executive briefings before they reach decision makers. There should be a strategy and process to review and compare historical scenarios to ensure consistent, unbiased and context-aware reasoning.
Data quality remains the foundation, but robust AI governance acts as the critical guardrail ensuring that interpretations are transparent, comprehensive and entirely free from bias so that leaders and decision makers can make the best decisions.
Srikana, I really like the idea of treating AI-generated interpretations as unverified drafts and requiring traceability back to the source data.
Your point about human-in-the-loop review is particularly important.
What I continue to wrestle with is scalability. As organizations increasingly use AI to generate summaries, analyses, recommendations, and executive briefings, there may come a point where humans can no longer validate every interpretation in detail.
At that stage, trust becomes a critical organizational capability. The question may shift from "Can we validate every interpretation?" to "How do we determine which interpretations require validation and which can be trusted?"
Thanks for the thoughtful perspective. Saving Changes...
Ohh this is such a great question Imran. Data quality is the foundation, but the quality and accuracy of AI interpretation is equally critical, even perfect data can lead to poor decisions if AI misinterprets trends, misses context or introduces bias. Particularly, the callout is also Biasness of data, when AI summarises the data, sometimes if the value of data for a particular topic is low, AI might overlook it. From PMO prespective it might be cruicial piece of information which might have changed your decision , based on the context that human has. Organisations need AI governance, traceability and most importantly human-in-the-loop validation to ensure AI-generated insights are transparent, explainable and grounded in source data. We should always be able to trace the source of data.
As AI and different LLM models are evolving this is an ever changing landscape. Ultimately, the goal is not just “right data” but “right data + right interpretation + right governance” to enable confident, informed decisions.
...
1 reply by Imran Afzal
Aug 04, 2026 9:05 AM
Imran Afzal
...
I agree completely that data quality, interpretation quality, and governance all matter. The question I've continued thinking about since writing this post is whether interpretation itself is the real bottleneck.
Organizations rarely act on a single AI interpretation. Decisions emerge from many people interpreting dashboards, AI summaries, recommendations, conversations, and each other. Two teams can begin with identical data and even identical AI outputs, yet arrive at different decisions because they construct different shared interpretations.
That makes me wonder whether the next governance challenge isn't simply ensuring high-quality AI interpretations, but ensuring organizations can develop shared interpretations before they execute. AI is making interpretation faster and more abundant. It doesn't automatically make it more aligned.
Saving Changes...
Imran AfzalAuthor| The Strategic PMOCary, NC, United States
Aug 04, 2026 7:47 AM
Replying to Shweta Suresh Naik
...
Ohh this is such a great question Imran. Data quality is the foundation, but the quality and accuracy of AI interpretation is equally critical, even perfect data can lead to poor decisions if AI misinterprets trends, misses context or introduces bias. Particularly, the callout is also Biasness of data, when AI summarises the data, sometimes if the value of data for a particular topic is low, AI might overlook it. From PMO prespective it might be cruicial piece of information which might have changed your decision , based on the context that human has. Organisations need AI governance, traceability and most importantly human-in-the-loop validation to ensure AI-generated insights are transparent, explainable and grounded in source data. We should always be able to trace the source of data.
As AI and different LLM models are evolving this is an ever changing landscape. Ultimately, the goal is not just “right data” but “right data + right interpretation + right governance” to enable confident, informed decisions.
I agree completely that data quality, interpretation quality, and governance all matter. The question I've continued thinking about since writing this post is whether interpretation itself is the real bottleneck.
Organizations rarely act on a single AI interpretation. Decisions emerge from many people interpreting dashboards, AI summaries, recommendations, conversations, and each other. Two teams can begin with identical data and even identical AI outputs, yet arrive at different decisions because they construct different shared interpretations.
That makes me wonder whether the next governance challenge isn't simply ensuring high-quality AI interpretations, but ensuring organizations can develop shared interpretations before they execute. AI is making interpretation faster and more abundant. It doesn't automatically make it more aligned. Saving Changes...
It's both, but the bigger gap right now is interpretation quality, since most teams check data sources but rarely check the logic AI used to summarize them. Example: two managers feed the same sales data into AI and get different risk scores because each prompt framed "risk" differently, even though the data was 100% accurate. So the real fix isn't just clean data; it's a human checking how the AI reached its conclusion before it reaches a leader's dashboard. Saving Changes...
Community Champion
