Where Can AI Give Delivery Leaders Time Back This Week?

Start with the work surrounding delivery: creating initiatives, maintaining alignment, and preparing the status updates that explain what everyone has been doing.
I have seen a project update of three to five sentences consume five hours of combined effort, with three people refining the wording. Collecting updates. Reconciling accounts. Reviewing the language. Getting everyone comfortable with what those sentences say.
Measure that effort in your own organization. It gives you a practical starting point for deciding where AI could return leadership capacity.
In When an Initiative Stalls, Who Owns the Recovery?, I discuss the sponsor’s responsibility to connect delivery evidence with decisions. Preparing that evidence is one opportunity for AI. The work begins much earlier.
Start where the initiative starts
An initiative usually begins with a problem, an intended outcome, and some form of alignment documentation. Perhaps a business requirements document (BRD), a product requirements document (PRD), or a technical proposal.
Then it grows roots and branches.
Requirements become Epics and Stories. Technical decisions emerge. Scope changes. Teams discover dependencies. Someone updates a presentation. Someone else maintains a spreadsheet.
Eventually, the work being delivered bears only a passing resemblance to the document that started it.
Traceability becomes a distant memory. So do the missed requirements, until someone asks why they were never delivered.
AI can help prepare consistent initiative drafts and compare proposed work against the original intent. Give it the problem, expected outcome, known constraints, and existing evidence. Ask it to identify missing information, competing assumptions, and questions requiring a decision.
Use the same fields across initiatives so leaders can compare their outcomes, scope, assumptions, and dependencies. You do not have to start from scratch. You do have to review what it produces.
Create the work, then generate the document
Here is where I would challenge the sequence.
Writing a BRD, translating it into Epics and Stories, and maintaining both creates repeated effort and opportunities for divergence. Separately authoring and maintaining the same requirements twice is wasted effort.
Develop the requirements directly through the Epics and Stories. Capture the business outcome, scope, acceptance criteria, dependencies, and relevant decisions. Link the supporting business context and technical documentation. Review the work with the people responsible for business, product, technology, and delivery alignment.
Once those Epics and Stories are sufficiently defined and agreed, generate the static document from that body of work. That means completing the definition, not waiting until delivery is finished.
Where funding, audit, or governance requires an approved BRD, generate it for that purpose and retain the approved version and its links to the underlying work. Record subsequent changes and route them through the required approval process.
The goal is one source of truth. Required documents become dated representations of agreed requirements, with traceability to the work and its changes.
That gives AI something useful to compare: what was agreed, what changed, and what is now being delivered.
Prepare meetings from the delivery evidence
Jira is the tracker in my examples because the organizations I have worked with over the past eight years have used it.
Getting the information needed to have a useful meeting can consume more time than the meeting itself. With reliable daily updates, a practical target to test is pulling current information 30 minutes before the meeting instead of asking everyone for a separate update. Measure whether that window leaves enough time to verify the evidence and identify the decisions needed.
In my own practice, I work within both ChatGPT and Claude, using connectors to update Jira as the work develops. I can ask the assistant to update the tracker from the work we have just completed. Jira is a primary source of information and feeds my Decision AIOS, the decision support part of the broader AI Operating System (AIOS) I am building at NBD.
That is a workflow I use today. The delivery briefing prompt below is a proposed starting point for applying the same principle to meeting preparation.
For an organization using Jira and Atlassian’s Rovo, the workflow I would start with is pulling the relevant delivery information into a draft briefing for review.
A useful starting prompt:
Prepare a delivery briefing for this initiative using the available Jira records. Identify changed commitments, unresolved dependencies, missing acceptance criteria, and decisions requiring leadership attention. Link each material statement to its source. Separate recorded facts from interpretations, and identify information you could not establish.
Use an approved environment, define access boundaries, and limit the workflow to authorized sources. Keep human review and approval for material statements, decisions, and communications, with the relevant operational and control owners involved.
The review then becomes more focused. Does the briefing accurately represent the evidence? What needs clarification? Which decisions need attention?
That is Traceable Interpretation: keeping a usable path from a conclusion back to the evidence and assumptions supporting it.
AI may expose the habits behind the data
Everyone worries that AI will give them incorrect information. Fair concern. Our reporting habits deserve the same scrutiny.
We can spend an impressive amount of time improving the wording of a status update. If only the dependency moved as quickly as the adjectives.
Delivery reporting already contains interpretation. Leaders need to see where the recorded evidence ends and someone’s explanation begins.
If commitments, blockers, and decisions live primarily in conversations, the reporting process depends on reconstructing those conversations later.
Start improving the evidence through four daily habits:
Update the work when a commitment changes.
Record a blocker when it emerges.
Capture a decision with its owner and rationale.
Maintain acceptance criteria as understanding develops.
Those habits improve the evidence available to both people and AI.
AI tools are taking meeting notes now, too. Review those notes, confirm the decisions and owners, and carry the relevant actions into the tracker. That gives the next meeting a record to work from.
Spend the preparation time checking facts and identifying decisions that need attention. In the spirit of Sergeant Friday: just the facts.
Connect preparation time to decision latency
This week, start by measuring preparation effort. Decision latency is the second step: begin recording timestamps now, then assess the pattern over subsequent delivery cadences.
For this workflow, I would measure Decision Latency as the elapsed time between identifying a decision that requires action and recording an authorized decision. Capture when it was raised, when sufficient evidence became available, and when it was made. This distinguishes evidence preparation from waiting for authority or attention.
Consider a dependency needing a sponsor’s decision on Tuesday but waiting for Friday’s status meeting. A briefing prepared in minutes has limited value if the decision still waits until Friday. Current evidence needs a route to the decision owner and an escalation point based on when delivery could be affected.
The cost of delay is the economic consequence of that wait: deferred business value, additional delivery expense, or rework. Ask what remains blocked, whether people can do useful alternative work, and whether the release date changes. Three days waiting does not automatically mean three days added to delivery. Estimate the consequence from the business case and delivery conditions, make assumptions visible, and avoid counting the same impact twice.
AI can help assemble evidence and surface aging decisions. The authorized leader still makes the tradeoff. Whether time reclaimed leads to earlier action is something to measure.
Measure the process you want to improve
Establish your baseline now.
How long does it take to create a BRD and translate it into Epics and Stories? How much effort goes into reconciling them afterward? How many combined hours produce the weekly status update?
Then measure the revised workflow, including review and correction time.
Track preparation effort, decision latency, and the delivery consequence separately. Compare the baseline with the revised workflow: did preparation take less effort, did the decision happen sooner, and did that change the outcome?
The useful outcome is reclaimed capacity with requirements still traceable, decisions visible, and less repeated preparation.
That is where AIOS fits: designing connected workflows with clear evidence, roles, review, and accountability. I explain the broader operating model in Building an AI-Native Company: From Experimentation to an Operating System.
Choose one this week. Initiative creation, requirements alignment, or meeting preparation. Measure the current effort, establish the workflow, and see what time you can give back.
Through Fractional Executive & Strategic Advisory (FESA), I help technology and delivery leaders improve how evidence becomes decisions and action. We can start by identifying one workflow, establishing its effort baseline, and defining the evidence and review needed to trial it. Discuss your delivery leadership priorities.
About the author
Duane L. Nicholson is Founder and Chief Transformation Officer of NBD Consulting Services and a Fractional Executive & Strategic Advisor. He works with technology, product, and delivery leaders on modernization, operating model change, and strategic execution. His experience includes DTCC, Experian, Amgen, Toyota Financial Services, and Pacific Life. He is a SAFe Practice Consultant (SPC).




Comments