The Readiness Divide in AI-Assisted Operations Consulting

Spend enough time inside transformation engagements and a pattern becomes impossible to ignore: the clients asking for AI-assisted operations in 2026 are divided less by budget or industry than by whether their people are prepared to supervise intelligent systems at all. The tooling conversation has become almost easy — detection, correlation and automated response capabilities are mature and widely available. What separates the engagements that compound value from the ones that quietly stall is the state of the team on the receiving end. This piece examines that divide as a trend, drawing on what consulting work across enterprises keeps surfacing, and argues that enablement has become the real deliverable.

The demand signal has changed shape

Three years ago, a typical brief read like a procurement exercise: evaluate intelligent monitoring platforms, recommend one, integrate it. The current generation of briefs reads differently. Clients arrive having already bought something — often two somethings — and the ask is why the promised reduction in manual toil never materialised. In discovery, the answer is rarely technical. The platform is flagging genuine irregularities in release behaviour, infrastructure load and job scheduling; the operations staff are treating those flags as yet another alert feed to acknowledge and move past. The organisation purchased a capability its people were never prepared to exploit.

This is the consulting market’s version of a well-documented adoption law: capability absorbed is a function of skills present, not licences owned. The pattern echoes what we have long argued about transformation generally — that outcomes follow operating-model change, a case we built in our strategy piece on consulting as the lever for enterprise agility — but the AI-assisted wave sharpens it, because the technology actively requires informed human judgement to calibrate.

Where client teams actually stand on AI-assisted operations

Across engagements, client operations teams tend to cluster into three observable groups. The first still runs on manual pattern-matching: experienced engineers who can smell a bad deploy, none of whom can articulate how a learned baseline differs from a static threshold. The second has isolated expertise — one architect or SRE who understands detection models, drift and automated remediation, and has become a single point of failure for the entire ambition. The third, still a minority, has distributed fluency: multiple people per rotation who can read, challenge and tune what the machines conclude. Only the third group extracts compounding value, because only there does the feedback loop between human correction and machine learning actually close.

The evidence for which group a client belongs to shows up in small artefacts. Ask to see the last ten anomaly flags and how each was resolved. Group one has no flags. Group two has flags resolved by one named individual. Group three has flags annotated by several hands, including overrules — the healthiest signal in the whole assessment, because it means humans are exercising judgement over the system rather than deferring to it.

Enablement is becoming a contract line item

The most telling trend of 2026 is commercial: structured upskilling is moving from a goodwill gesture at the end of an engagement into a priced, scheduled workstream at the start of one. Statements of work increasingly specify who on the client side will study what, by when, and how competence will be evidenced. External curricula make this contractible in a way that ad-hoc knowledge transfer never was. AIOps School’s progression of four certification tiers — running from fundamentals up through engineering, strategy and architecture levels — gives an engagement a vocabulary for saying “two platform engineers to the hands-on engineering tier, the operations manager to fundamentals” and attaching dates. Its entry programme is a month of part-time study; the deeper technical track, which takes in detection modelling and self-healing design, spans roughly a month and a half. Those durations fit inside a transformation timeline, which is precisely why they are being written into them.

The same logic explains a second commercial shift: clients asking for named-person readiness plans rather than team-level training budgets. A budget disperses; a plan that says which individual reaches which competence by which sprint is auditable. Consultancies that resist this granularity are finding that clients now request it unprompted — a reversal from even two years ago.

What the stalled engagements have in common

The negative evidence is equally instructive. Reviewing engagements that failed to produce durable automation, three antecedents recur. First, the client automated responses before its team trusted the detections, so the first false-positive remediation froze the programme politically. Second, knowledge was transferred to whoever attended the workshops rather than to the roles that would hold the pager, leaving the night shift exactly where it started. Third, no one owned the chore inventory — the ranked list of repetitive operational work that tells you what to hand to machines next — so automation proceeded by vendor demo rather than by value. All three are organisational failures that a disciplined engagement structure prevents, which is why we keep returning to the sequencing described in our blueprint for running an engagement from first assessment to client enablement.

The forecast: readiness as the differentiator

Projecting the current trajectory forward, three developments look probable. Client-side job descriptions for operations roles will routinely list anomaly-detection literacy the way they once listed scripting. Engagement pricing will continue shifting from platform integration toward person-by-person capability building, since that is where the unrealised value now sits. And the consultancies themselves will be held to the same bar — a consultant who cannot demonstrate hands-on, lab-tested familiarity with the systems they recommend will struggle in rooms where the client has already put staff through structured cohorts. It is notable that training providers in this space now also operate practitioner marketplaces and advisory arms; the boundary between learning and delivery is dissolving from both sides.

For leaders planning 2027 budgets, the implication is uncomfortable but actionable: the constraint on AI-assisted operations is no longer available technology, and it has not been for some time. It is the distribution of understanding across the people who run production. Treat that distribution as the primary asset the engagement exists to build — measured, named and scheduled — and the platforms will finally pay for themselves. Treat it as an afterthought, and 2027’s briefs will read exactly like 2026’s: why did the toil never go down?

Kabir Malhotra