FUTURECENTRAL PRESS · BOOK SAMPLE
The Chief AI Officer
Building and Working in the Enterprise AI Function
Chapter 3: The First 90 Days of an AI Mandate
Winning the quarter with mandate clarity and one defensible win, not a strategy deck
Chapter Outcomes
Chapter 2 drew the map of a working function; this chapter faces the ninety days in which one must be stood up. It answers what anyone handed an AI mandate, or supporting someone who was, needs settled in the first quarter, before the strategy documents and the org-design debates consume the only window in which a new function is judged on momentum rather than results.
First, how to convert an ambiguous mandate into a defensible scope that the sponsor will recognize and the organization cannot quietly renegotiate later.
Second, how to choose a first win that is visible enough to matter, winnable inside the quarter, and substantial enough that no one can dismiss it as a toy.
Third, how to triage the first quarter’s work into what to build, what to borrow, and what to defer, so that scarce credibility and scarcer engineering time are spent where they compound.
Fourth, how to set the reporting rhythm deliberately, before the organization sets it by default, because the cadence a new function establishes in its first quarter is the cadence it will be held to for years.
Opening Vignette
Kavita Rao was given a sentence and a deadline nobody stated. The sentence, delivered by the managing director of a mid-size Indian life insurer at the end of a strategy offsite, was "make us an AI company." The deadline was the ninety-day informal probation that every newly created leadership mandate carries in an organization that has not done this before, and that no one ever writes down.1 She had been the transformation lead for the digital-channels program, which is why the mandate landed on her, and she understood the sentence perfectly and the assignment not at all.
The insurer already had AI in the building, scattered and unsponsored. The actuarial team ran predictive models that predated the phrase; the claims unit had piloted a fraud-scoring tool; the distribution team wanted a generative assistant for its agents; and a vendor had left behind a proof-of-concept chatbot that worked in the demo and nowhere else. Kavita had no team, no budget line she controlled, and a title that made her responsible for all of it and accountable for none of it in any way the organization had defined.2
What she did have, and would only have once, was a quarter of unearned credibility. The organization was willing, for about ninety days, to treat her as the person who would make AI real, and to extend her the benefit of the doubt that a new mandate enjoys before it has been measured. That credibility was the scarcest asset she would ever hold, and it was depreciating from the day of the offsite. She could spend it on a strategy deck that would impress the board and change nothing, or on a single shipped result that would convert the goodwill into a mandate with a budget behind it.
The decision this chapter examines: with one quarter, no team, and a mandate written as a slogan, where does a new AI leader spend the only credibility they will ever be given for free, and what single result is worth staking that quarter on?
3.1 Converting an ambiguous mandate into a defensible scope
The first task of a new AI mandate is not to plan the work but to bound it, because an unbounded mandate is less a grant of authority than an invitation to be blamed for everything. "Make us an AI company" is a direction, and a direction is not a scope; the leader who accepts it as written has agreed to be accountable for an outcome they cannot define, deliver, or defend.
The organization that handed down a slogan will, when something goes wrong, interpret the slogan to cover whatever failed, and the leader who never bounded the mandate has no basis to decline the blame. Converting the mandate means writing down, and getting the sponsor to accept, what the function owns, what it influences, and what it explicitly does not touch this quarter. This is the mandate-clarification work of Chapter 1, now executed under a clock. An ambiguous mandate must be converted into a defensible scope in the first weeks, because a mandate that stays ambiguous defaults to the broadest possible reading at the worst possible moment.
Kavita’s actuarial models, the claims fraud pilot, and the abandoned chatbot are not distractions from her mandate; they are her mandate’s raw material. A new leader who maps the existing AI, names an owner for each piece, and identifies the one gap worth closing first has converted a slogan into a portfolio in a fortnight. The leader who instead opens with a twelve-month strategy has spent the first weeks producing a document that the second month will invalidate. The defensible scope is built from the work already underway, not from a blank sheet, because the fastest route to a genuine mandate is to organize what exists rather than to invent what does not.
The managing director who said "make us an AI company" has a private picture of what success looks like in ninety days, and that picture, not the slogan, is what the leader will be judged against. Surfacing it, by proposing a concrete first-quarter scope and watching where the sponsor pushes back, is how the leader learns the actual terms while there is still time to meet them. The sponsor conversation deferred is the probation failed in slow motion. The scope conversation with the sponsor is the one that must happen early and explicitly, because the sponsor’s unstated expectations are the genuine terms of the probation.
Every function around the new mandate will ask it to take on their AI problem, and each request will be reasonable in isolation. The leader who says yes to all of them has rebuilt the unbounded mandate by accretion and guaranteed that the quarter ends with much motion and no result. Narrowing the scope is not timidity; it is the precondition for shipping anything at all, and Section 3.2 turns to the single result the narrowed scope should aim at. A defensible scope is narrow on purpose, and the discipline to keep it narrow under pressure to promise everything is the first test of a new AI leader.
3.2 Choosing the first win — visible, winnable, not a toy
The first quarter is won or lost on a single decision: which result to ship. Everything else in the first ninety days, the scoping, the triage, the cadence, exists to serve that one choice, because a new function that ends its first quarter with a shipped, credible result earns a second quarter with a budget, and one that ends with a strategy deck earns a review.
It must be visible, meaning its success registers with the people who decide the function’s future. It must be winnable inside the quarter, meaning the data, the sponsor, and the path to production actually exist. And it must not be a toy, meaning it addresses a genuine cost, risk, or revenue line rather than a demonstration that impresses in a review and changes nothing. The three conditions are a filter, and most of the projects a new leader is offered fail at least one of them. The first win must satisfy three conditions at once, and a candidate that fails any one of them is the wrong first project regardless of how attractive it looks.
A modest model that removes a visible, chronic pain the business complains about weekly is worth more in the first quarter than a sophisticated system whose value is real but illegible to the people funding the function. The claims-fraud score that a chief financial officer can read as recovered rupees is a better first win than an elegant pricing model whose benefit takes three slides to explain. Visibility is chosen, and choosing it is a political act the new leader should make deliberately. Visibility is about the audience for the result, not the sophistication of the model, and the first win should be chosen for who will notice it rather than for how advanced it is.
The evidence on enterprise AI is unforgiving here. Roughly ninety-five percent of enterprise generative-AI pilots deliver no measurable return by the most widely cited finding, and about eighty percent of AI projects fail to deliver their intended business value by the broader estimate. Both failures trace largely to projects that were never winnable inside their window, because the data, the integration, or the ownership was missing from the start.3 A winnable first project is one where the data is present and clean enough, the sponsor is real, and the path to production is short. The leader who confirms these before committing has already avoided the failure mode that consumes most first quarters. Winnability is about the honest state of the data and the sponsorship, and the most common first-quarter failure is to pick a project whose data or approvals do not actually exist.
A chatbot that answers questions in a demo, a dashboard that visualizes what everyone already knows, a generative tool that drafts text no one was struggling to write: these ship quickly and prove nothing, and the organization can tell. A first win that is winnable and visible but trivial spends the quarter’s credibility on a result that invites the question the function cannot afford, which is whether any of this was worth a dedicated leader. The first win should be small in scope and serious in stakes, and Section 3.3 turns to how to resource it. The not-a-toy test is the one new leaders most frequently fail, because a toy is the easiest thing to ship and the most tempting thing to show.
3.3 Build, borrow, defer — the first-quarter triage
A new function cannot build everything the mandate implies, and the attempt to do so is how first quarters are lost. The triage that separates what to build from what to borrow and what to defer is the discipline that enables a mandate with no team and no budget to ship a genuine result inside ninety days, and it is a triage the new leader has to run explicitly rather than by drift.
The first win needs one model, one integration, or one workflow built to production quality, and the leader’s scarce engineering time should concentrate there. Every other build the organization proposes in the first quarter, however worthy, competes for the same hours and dilutes the one result that matters. Building narrow rests on the recognition that a single shipped result is worth more than five half-built ones. Build only the one thing the first win genuinely requires, and treat every additional build request as a threat to the quarter rather than an addition to it.
The platform can be a managed cloud service rather than a built one; the model can be a vendor’s or an open-weight one rather than a bespoke build; the data pipeline can reuse what the analytics team already runs. Borrowing is the correct default for everything that is not the differentiating core of the first win, and the leader who insists on building the undifferentiated plumbing has chosen to feel in control at the cost of the quarter. Chapter 6 develops the build-versus-buy discipline in full; in the first quarter, the bias toward borrowing is close to absolute. Borrow the capability that already exists rather than rebuilding it, because the first quarter has no time to construct what can be acquired or reused.
The distribution team’s generative assistant and the actuarial team’s model refresh are not rejected; they are scheduled for after the first win ships, with a named quarter attached. Deferral without a date is a brush-off that breeds resentment; deferral with a date is a roadmap that buys patience. The leader who defers well keeps the goodwill of the functions they are not serving yet, which is the goodwill they will need when the first win succeeds and the second wave begins. Defer, visibly and with a date, everything that is not the first win, because deferral with a commitment is how a leader says no without making an enemy.
A function that must borrow its platform because it has no platform spine, and borrow its governance because the second line is not yet staffed, is reading its own role map correctly. The build-borrow-defer triage and the dependency-ordered role map are the same discipline seen from two angles: build where the function has the role, borrow where it does not yet, and defer what requires a role the function has not built. Section 3.4 turns to the rhythm that holds the quarter together. The triage compounds with the role-building sequence of Chapter 2, because what a new function builds first should match the roles it can staff first.
3.4 Setting the reporting rhythm before it sets itself
The least glamorous decision of the first quarter is the one that most shapes the function’s future: how, to whom, and how frequently the function reports. A new leader who does not set the cadence deliberately will have it set for them, usually by whichever executive asks for an update first, and the improvised cadence that results is almost always wrong in a way that is difficult to correct later.
A function that begins by reporting model metrics will be asked for model metrics forever; a function that begins by reporting business outcomes trains its audience to expect business outcomes. The first few updates a new leader provides are not status reports but the setting of a standard, and the leader who understands this chooses the terms of the first update as carefully as the first win itself. The reporting rhythm should be established in the first weeks, because the cadence a function sets early is the cadence it is held to for years.
A managing director does not want the area under the curve; they want the recovered rupees, the reduced cycle time, the risk removed. The leader who reports in model terms forces the sponsor to do the translation and, in doing so, quietly cedes control of how the function’s value is judged. Chapter 10 develops the full value-attribution discipline; in the first quarter, the rule is simply that the update speaks the funder’s language from the first sentence. The cadence must report to the sponsor in the sponsor’s language, which is almost never the language of models, and the translation is the leader’s job rather than the audience’s.
A crisp fortnightly note that shows the first win advancing builds more confidence than a monthly deck that shows activity, because momentum is the thing a new function most needs to demonstrate and the thing a slogan-mandate was given to produce. The public-sector experience is instructive: the United States federal AI mandate paired the designation of each agency’s AI leader with a fixed reporting and inventory cadence rather than leaving the rhythm to chance, on the logic that a new accountability holds only if its reporting is regular.4 The cadence is set once and lived with for years, so it is worth setting on purpose. The framework that follows turns the first-win choice into a scored decision. The rhythm should be frequent enough to show momentum and infrequent enough to have something to say, and for a first quarter that usually means a short, regular update rather than a rare, elaborate one.
Framework: The First-Win Selection Grid
The First-Win Selection Grid is the tool that turns the most consequential decision of the first quarter, which single result to ship, from an argument into a scored comparison. It rates each candidate first project on three axes and forces the trade-offs into the open, so that the leader stakes the quarter on the project the evidence supports rather than the one the loudest sponsor prefers. Its output is a single choice the leader can defend.
A candidate scores high on visibility when its success is legible to the sponsor and the funders without translation, and low when its value is real but takes three slides to explain. Visibility is scored from the funder’s seat, not the builder’s, and a project that the engineering team finds fascinating but the chief financial officer cannot read scores low no matter how elegant it is. The first axis is visibility: how clearly the result will register with the people who decide the function’s future.
A candidate scores high on winnability when the data is present and clean enough, the sponsor is real and engaged, and the route to production is short and unblocked. It scores low when any of these is aspirational, and the discipline of the grid is to score winnability honestly rather than optimistically, because the optimistic score is the one that produces the pilot that never ships. The second axis is winnability: how honestly the data, the sponsorship, and the path to production already exist.
A candidate scores high on strategic signal when success makes the case for the next investment obvious, and low when it is a one-off that proves nothing beyond itself. The strategic-signal axis is what separates a first win that funds a second quarter from a first win that is merely a completed task, and it is the axis that keeps the grid from selecting a safe triviality. The third axis is strategic signal: how well the result, once shipped, argues for the function’s continued existence and expansion.
The leader lists the candidate first projects, scores each on visibility, winnability, and strategic signal, and stakes the quarter on the candidate that scores well on all three rather than the one that scores highest on any single axis. A project that is visible and strategic but not winnable is a trap; a project that is winnable and visible but sends no strategic signal is a wasted quarter. The grid closes on the question the chapter opened with, which single project is worth the only free credibility the leader will ever have, and the Applied Exercise builds it on the reader’s own candidates. The grid’s use, for the leader running a first quarter, is that it converts a contested choice into a defensible one by making the three axes explicit and scoring every candidate on all of them.
India Cases
Case 1, the Indian bank that made fraud and collections its visible first win. The strongest first-quarter wins in Indian banking have come in fraud and collections, because they move a number the board already watches. India’s largest private-sector banks have built machine learning into exactly these workflows: HDFC Bank runs model-driven transaction monitoring inside its fraud operation, and ICICI Bank has embedded machine-learning scoring in its credit-risk and early-warning systems, deployments both banks describe in their own disclosures.5 The precise gains are not publicly audited, and the percentages that circulate in trade coverage are directional at best; what is verifiable is that both banks have kept funding and widening these systems year after year, which is the revealed-preference evidence that the deployments earn their keep. Read for the first-win logic, these deployments show why a fraud or collections model is the archetypal first win: it is visible to the chief financial officer as recovered rupees or avoided losses, it is winnable because the transaction data is already clean and abundant, and it sends a strong strategic signal because it moves a board-level number. The first-time AI lead choosing a first win from a field that also includes a customer-service chatbot should note which candidates can shift a figure the board already tracks, and which can only promise a demonstration.
Scenario 2, the enterprise that spent its first quarter on a strategy and lost the mandate. The counter-example is the more common one, and it is worth stating plainly because it is the outcome this chapter exists to prevent. An Indian enterprise that stood up an AI function and spent the first ninety days producing an ambitious multi-year AI strategy, with workstreams, target operating models, and a transformation roadmap, entered its second quarter with an impressive document and nothing shipped.6 The strategy was not wrong; it was premature, because it asked the organization to keep believing on the strength of a promise at exactly the moment the organization wanted proof. When the enthusiasm faded, as it does around the second quarter, the function had spent its free credibility on a deck and had no shipped result to renew it. The mandate thinned, the budget conversation went badly, and the function devoted its second year to recovering the credibility its first quarter had spent on the wrong artifact.
Global Cases
Case 3, the documented enterprise playbook of shipping a narrow win first. The most durable finding from the study of enterprises that have deployed AI successfully is that the ones that scaled started with a narrow, high-leverage workflow, shipped something useful in weeks, and used the early win to fund the next, while the ones that failed started with a moonshot.7 The Stanford Digital Economy Lab’s review of fifty-one successful enterprise deployments makes the same point from the winning side that the failure statistics make from the losing side: a genuine quick win deploys in under ninety days, targets one workflow, and produces a number a chief financial officer recognizes. Read for the durable logic beneath the specifics, the lesson for an Indian AI leader is that the sequence, narrow win first and expansion second, is a structural property of how enterprises come to trust a new function rather than a national preference.
Case 4, the public-sector mandate that built cadence into the role. The United States federal government, in mandating a Chief AI Officer for every agency, did not leave the reporting rhythm to each appointee’s discretion; it paired the role with a fixed cadence of compliance plans and public AI-use inventories on a defined schedule.8 Read against this chapter’s argument, the federal design encodes the discipline that a new accountability holds only if its reporting is regular and its scope is inventoried rather than assumed. What transfers to the enterprise AI leader is the principle that cadence is part of the mandate, not an afterthought to it. What an enterprise leader must supply that the federal design assumed is the first shipped win, because a public mandate is backed by statute while an enterprise mandate is backed only by the credibility the first quarter earns.
Practitioner’s Lens: The transformation lead on an unstated ninety-day probation
The transformation lead handed an AI mandate spends the first quarter managing a probation that no one has admitted exists, and the skill that carries her through it is the discipline to convert goodwill into a shipped result before the goodwill expires.9 She was chosen because she had run a transformation program before, which means the organization trusts her to make something real, and that trust is both her asset and the thing she is about to spend. Her first quarter is a single sustained act of choosing where to spend it.
A strategy is what a new leader is expected to deliver, and delivering it feels like progress, which is exactly why it is dangerous. The lead who produces a polished roadmap has met the expectation and shipped nothing, and she has trained the organization to wait for the next document rather than the next result. The more difficult and correct move is to deliver a narrow shipped win and let it stand in for the strategy the organization thinks it wants. The transformation lead’s first move is to resist the strategy deck that everyone expects her to produce, because the deck is how the quarter is lost most respectably.
She has seen data that was promised and did not exist, sponsors who were enthusiastic in the meeting and absent in the work, and integrations that were "nearly done" for a year. That scar tissue is her advantage: she scores winnability honestly where a first-time technologist scores it optimistically, and honest winnability scoring is what keeps her out of the ninety-five percent of pilots that never show a measurable return. The transformation lead’s sharpest judgment is in reading which of the candidate first wins is actually winnable, because her transformation experience makes her the rare new leader who can tell a genuine project from a demo.
Every function wants its AI problem to be the first win, and the lead can serve only one. The lead who defers the others with a named date and a genuine commitment keeps them as future allies; the lead who defers with a vague "later" converts them into quiet opponents who will not mourn the function’s struggles. Managing the deferred is as consequential as delivering the chosen, and the transformation lead’s instinct for organizational coalitions is what makes her good at it. The transformation lead’s most political act is the deferral, because saying no to the functions she is not serving yet, without making enemies of them, is what preserves the coalition she will need next quarter.
The slogan gave her a mandate with no substance; the first win gives the mandate a result the organization can point to; the kept deferrals give it a coalition; the deliberate cadence gives it a rhythm the organization can rely on. None of this is the grand strategy the offsite seemed to call for, and all of it is what actually turns a sentence spoken at an offsite into a function with a budget and a future. What the transformation lead is really doing in the first ninety days is converting a slogan into an institution, one shipped result and one kept promise at a time.
Applied Exercise: The One-Page 90-Day Plan
The chapter’s output is a one-page ninety-day plan the reader could take to a sponsor this week: a mandate statement, a first win, an explicit list of what is deferred, and a reporting cadence. It is a working document, not a strategy deck, and its one-page limit is the discipline. The exercise takes about two hours for a reader working on their own mandate, and it consumes the mandate-clarification memo from Chapter 1 as its first input.
Step 1, the mandate statement (twenty-five minutes). Write the bounded scope in three sentences: what the function owns this quarter, what it influences, and what it explicitly does not touch. This is the Chapter 1 mandate memo compressed to a scope a sponsor can accept or amend in one reading, and its narrowness is the point.
Step 2, the first win, scored on the grid (forty minutes). List the candidate first projects and score each on visibility, winnability, and strategic signal. Choose the one that scores well on all three, and write a single paragraph stating the win, the number it will move, and the date it ships. If no candidate scores well on all three, the honest output is that the first task is to create one, not to launch a weak candidate.
Step 3, the build-borrow-defer list (thirty minutes). For the chosen first win, identify the one thing to build, the capabilities to borrow rather than build, and the projects to defer with a named quarter attached to each deferral. The deferral dates are not decoration; they are the commitments that keep the un-served functions as allies.
Step 4, the reporting cadence (twenty-five minutes). Identify the sponsor, the frequency, and the language of the update, and draft the first update’s headline as it would read if the first win shipped on schedule. Writing the headline now, in the funder’s language, sets the standard the function will be held to and confirms that the chosen first win is one the sponsor will actually recognize as a win.
Executive Briefing
A new AI mandate is judged, for about ninety days, on momentum rather than results, and the leader who spends that window on a roadmap has converted the goodwill into a document instead of a result. The step to take this week is to identify the single first win worth staking the quarter on, and to stop work on everything that does not serve it. What comes first is to stop treating the first quarter as the time to produce a strategy and start treating it as the time to ship one defensible result.
Every attractive addition to the first win extends its timeline and lowers its odds of shipping inside the quarter, and a shipped narrow result beats an unshipped ambitious one every time. The discipline is to hold the first win small in scope and serious in stakes, and to route every "while you are at it" into the deferred list with a date. The first win the reader is most tempted to over-scope is the one to discipline hardest, because the instinct to make the first result impressive is what makes it un-shippable.
The multi-year AI strategy is not wrong, only premature, and presenting it in the first quarter spends credibility on a promise at the moment the organization wants proof. The roadmap lands very differently in the second quarter, when it can be introduced on the strength of a result already delivered rather than a result still hoped for. The roadmap the reader is being asked to present is the one to withhold until there is a shipped result behind it, because a strategy presented before a win is a promise the organization will stop believing.
A function that begins by reporting business outcomes in the funder’s language trains its audience to expect them; a function that begins with model metrics or with irregular, elaborate updates trains its audience to the wrong standard. The move is to choose the sponsor, the frequency, and the language of the update deliberately, and to make the first update a demonstration of momentum rather than a display of activity. The cadence conversation is the one to have this week, before the organization sets the rhythm by default, because the reporting terms a function establishes early are the terms it lives with for years.
The functions whose AI problems are not the first win will either wait as supporters or sour into critics, and the difference is whether their deferral came with a genuine date or a vague "later." The leader who spends the first quarter shipping one win and keeping several dated promises ends it with a funded mandate and a coalition, which is the position from which the next chapter’s question, where the function should sit and how its boundaries are drawn, can finally be answered. The deferred projects are the coalition the reader will need next quarter, and how they are deferred determines whether they become allies or opponents.
Key Terms
First-quarter mandate. The informal ninety-day window in which a newly created AI function is judged on momentum rather than results, and during which its founding leader holds a stock of unearned credibility that depreciates from day one.
Defensible scope. A bounded statement of what the function owns, influences, and explicitly does not touch this quarter, accepted by the sponsor. The conversion of an ambiguous mandate into terms the organization cannot quietly renegotiate later.
First win. The single shipped result a new function stakes its first quarter on. To qualify it must be visible to the funders, winnable inside the quarter, and substantial enough not to be dismissed as a toy.
First-Win Selection Grid. The chapter’s framework for choosing the first win by scoring each candidate project on three axes, visibility, winnability, and strategic signal, and selecting the candidate that scores well on all three rather than highest on any one.
Visibility. The degree to which a result registers with the people who decide the function’s future, scored from the funder’s seat rather than the builder’s. A result the sponsor cannot read without translation scores low regardless of technical merit.
Winnability. The honest state of the data, sponsorship, and path to production for a candidate first win. Optimistic winnability scoring is the leading cause of pilots that never reach production.
Strategic signal. The degree to which a shipped result argues for the function’s continued existence and expansion. The axis that separates a first win that funds a second quarter from a completed task that proves nothing beyond itself.
Build-borrow-defer triage. The first-quarter discipline of building only what the first win genuinely requires, borrowing every undifferentiated capability, and deferring everything else with a named date attached.
Deferral with a date. The practice of declining a project by scheduling it for a named future quarter rather than an open-ended "later." The mechanism by which a leader says no to un-served functions while keeping them as allies.
Reporting cadence. The frequency, audience, and language of a function’s updates, set deliberately in the first weeks. The cadence a function establishes early is the standard it is held to for years, which is why it must be chosen rather than inherited.
Reflection Questions
Answering these usefully assumes the reader holds, or expects to hold, an AI mandate in their own organization, and wants to apply the chapter to their own first quarter rather than discuss it in the abstract.
1. Write the bounded scope of your own AI mandate in three sentences: what you own this quarter, what you influence, and what you explicitly do not touch. Where would your sponsor push back, and what does that push-back tell you about the genuine terms of your mandate?
2. List the candidate first wins available in your organization and score each on visibility, winnability, and strategic signal. Which scores well on all three? If none does, what is the smallest project you could create that would?
3. For your chosen first win, run the build-borrow-defer triage: what is the one thing you must build, what can you borrow, and what will you defer? Attach a named quarter to each deferral, and identify which deferred function is most likely to become an opponent if handled badly.
4. Consider the transformation lead in the Practitioner’s Lens. Are you being pulled toward producing a strategy deck in your first quarter? What would you have to ship instead to make the deck unnecessary, and what is stopping you from choosing the shipped result over the document?
5. Draft the headline of your first reporting update as it would read if your first win shipped on schedule, written in your sponsor’s language rather than in model terms. Does the headline describe a number your sponsor already cares about? If not, is the first win visible enough?
6. The chapter argues that the cadence a function sets early is the cadence it is held to for years. What reporting rhythm is your organization about to set for your function by default, and what would you set deliberately instead before the default takes hold?
Further Reading
For the first-quarter discipline itself, the reading that helps most is the growing body of enterprise-AI deployment playbooks together with the empirical studies of why pilots fail to scale, taken as two views of the same finding. The Stanford Digital Economy Lab’s review of successful enterprise deployments and the MIT-affiliated research on the fate of enterprise generative-AI pilots are the two anchors, and they are most valuable read as a pair, because the success playbook and the failure statistics describe the same narrow-win-first logic from opposite ends.
For the transition-into-a-new-mandate craft that this chapter applies to the AI function specifically, the standard reference remains the general management literature on the first ninety days of any new leadership role, which supplies the underlying logic of early wins, coalition-building, and cadence that this chapter translates into AI terms. The working professional stepping into an AI mandate should read that general treatment alongside this chapter, because the AI-specific traps sit on top of the universal ones rather than replacing them.
For the Indian enterprise context, the NASSCOM and industry-body reports on Indian enterprise AI adoption, and the trade-press coverage of Indian firms standing up AI functions through 2024 and 2025, are the working reference for what first quarters actually look like in the Indian market, where the talent, data, and sponsorship conditions differ from the global playbooks in ways the leader has to read locally. These sources are best used to calibrate the global playbook to Indian data-readiness and sponsorship realities rather than as standalone guidance.
Chapter 4 takes up where the function should sit and how its boundaries are drawn, shifting from the first quarter’s momentum to the structural question of authority. The reader who has produced the one-page ninety-day plan from this chapter’s Applied Exercise will read Chapter 4 with the first win already chosen, which is the right position from which to negotiate the reporting lines and boundaries that determine whether the function has the authority to deliver it.
Endnotes
Kavita Rao is a composite character constructed from publicly reported patterns in the experience of first-time enterprise AI leaders at mid-size Indian firms during 2023–2026. Her specific circumstances are illustrative and do not describe any single individual or firm.
The pattern of scattered, unsponsored AI inside an enterprise before a dedicated function is created is documented across enterprise-AI adoption studies, 2023–2026; the insurer’s specific inventory here is illustrative.
For the finding that a large majority of enterprise generative-AI pilots fail to reach production or show measurable business impact, see the MIT Project NANDA report, The GenAI Divide: State of AI in Business 2025 (2025), widely reported as the ninety-five-percent figure; for the broader estimate that roughly eighty percent of AI projects fail to deliver intended business value, see RAND Corporation, The Root Causes of Failure for Artificial Intelligence Projects and How They Can Succeed (RR-A2680-1, 2024), and the related analyst literature. Both figures are directional and vary by definition and sample.
For the United States federal Chief AI Officer mandate and its fixed reporting and inventory cadence, see Office of Management and Budget Memorandum M-24-10 (March 28, 2024), which paired the CAIO designation with compliance-plan and AI-use-inventory obligations on a defined schedule; OMB replaced M-24-10 with Memorandum M-25-21 in April 2025, retaining both the designation and the inventory obligation.
For HDFC Bank’s machine-learning-based transaction monitoring and fraud controls and ICICI Bank’s machine-learning-based credit-risk and early-warning systems, see the banks’ annual reports and investor disclosures, 2023–2025. Specific percentage improvements circulate in trade coverage but are not disclosed as audited figures by either bank; the case is cited for the type of first win a fraud or collections deployment represents, not for particular numbers.
This scenario is a composite drawn from publicly reported patterns in enterprises that prioritized strategy production over shipped results in a new AI function’s first quarter; specific details are illustrative.
For the finding that enterprises which scaled AI started with narrow, high-leverage workflows and used early wins to fund expansion, while those that failed began with moonshots, see the enterprise-AI deployment playbook literature, 2025–2026, including the Stanford Digital Economy Lab’s The Enterprise AI Playbook: Lessons from 51 Successful Deployments (Pereira, Graylin, and Brynjolfsson, 2026).
For the federal cadence design, see OMB M-24-10 (2024) and the associated compliance-plan and inventory schedule, carried forward in its successor M-25-21 (2025); cited here as a representative public-sector benchmark for building reporting rhythm into a mandate.
The Practitioner’s Lens is a composite constructed from publicly reported patterns in the experience of transformation leaders handed AI mandates; it is a teaching device, not an account of a specific individual or firm.