Adoption, Not Activation
Turning an agent on isn't the same as a customer getting value. If we don't measure adoption, customers renew less and advocate less - and we lose the credibility to sell the next agent.
Welcome
An agent being live in a tenant is the start of the job, not the end of it. Too many programmes stop the clock at go-live, log the activation, and move on - then wonder why usage is flat at month three and the renewal conversation goes sideways.
This module is for the people who own what happens after the switch flips. Customer Success leads it; delivery and engagement managers need to understand it well enough to set the right expectations from day one. By the end, you'll have a defensible cadence, a short list of metrics worth tracking, and a playbook for when the numbers go the wrong way.
Learning objectives
Activation vs adoption
These two words get used interchangeably. They shouldn't be. The commercial gap between them is where renewal risk lives.
Activation
The agent is switched on in the tenant, permissions are in place, the pilot user group can log in. A configuration milestone.
Adoption
Real users use the agent in their real jobs, often enough to change the outcome. A behaviour and value milestone.
The gap
Activation is binary and takes days. Adoption is gradual and takes months. Mistaking one for the other is how deals stall at renewal.
Concrete example: a customer activates the Recruiting agent in week two of go-live. Two months later, three recruiters use it daily, eleven ignore it, and two actively override its recommendations. Activation: done. Adoption: at risk. The customer's executive sponsor sees "rolled out" on the status report and a flat hiring cycle time - and starts to wonder what they paid for.
Our job isn't to deliver an activation - anyone can do that. Our job is to deliver adoption, because adoption is what the customer renews on and what earns us the right to sell the next agent.
Leading vs lagging metrics
Track both. Leading metrics tell you what's happening right now so you can intervene. Lagging metrics tell you whether it worked so you can prove value. A programme that reports only lagging metrics finds out too late; one that reports only leading metrics has no story at renewal.
Leading - steer with these
Usage frequency - invocations per active user per week.
Active users - % of licensed population actually invoking the agent.
Agent invocations - total volume, broken down by team and workflow.
Override rate - how often users reject or redo the agent's output.
Task completion time - time to finish the job with the agent vs without.
Lagging - prove value with these
Business outcomes - time-to-hire, forecast accuracy, cost saved per process, cycle-time reduction.
Renewal - did the customer renew, and at what size?
Expansion - did they buy the next agent or widen the footprint?
NPS / sentiment - what do users and the exec sponsor say about the programme?
Rule of thumb: every QBR deck should pair at least one leading metric with one lagging metric. Leading-only looks like activity; lagging-only looks like a report card with nothing to act on.
Health-check cadence
Adoption doesn't happen on its own. The cadence below is the minimum viable rhythm - every engagement should meet it, and the most successful ones do more in the first month, not less.
Weekly · Month 1
Usage dashboard review, pilot-user stand-up, blocker list, quick-fix configuration tweaks, early sentiment read.
Monthly · Months 2-6
Leading-metric trend review with customer lead, override analysis, training top-ups, expansion-team onboarding, first lagging signals.
Quarterly · 6 months+
QBR with exec sponsor - leading and lagging metrics, business-outcome narrative, next-agent roadmap, renewal signal check.
What's covered each time
- Weekly (month 1): short, operational, focused on removing friction. Who logged in, who didn't, what tripped them up, what needs a config change by Friday.
- Monthly (months 2-6): tactical, trend-focused. Is usage widening beyond the pilot team? Is the override rate falling as users trust the agent? What does task-completion time say?
- Quarterly (6 months+): commercial and strategic. Business outcomes tied back to the original business case, NPS, expansion conversation, input to the next cycle of the Kainos AI Navigator.
Going straight to quarterly in month one. Every programme that has renewed badly skipped the weekly rhythm. The first month is where adoption is won or lost.
Red flags
Five patterns show up again and again in programmes that are about to stall. Any one of them on its own is a conversation; two together is an intervention.
Intervention playbook
A red flag isn't a reason to panic - it's a reason to run the right play. Expand each one for the move.
Flat usage after month 1
First question: is it discovery or motivation? Interview five non-users - if they don't know how to get to the agent, it's a comms and enablement fix (in-product nudges, manager briefings, refreshed training). If they know and choose not to, the agent isn't solving a problem they care about; re-scope the use case or move on to a team that does have the pain.
Spike in override rate
Pull a sample of overrides and look for a pattern. Common causes: agent trained on incomplete data, edge cases the configuration missed, or users working around a quirk rather than reporting it. Fix the underlying issue - a configuration tweak, a data-quality job, or a clarifying piece of training - and track override rate weekly until it stabilises.
Sentiment drops
Don't wait for the survey result to be "final". Run a short qualitative session with the pilot team and one adjacent team, separate the "agent is wrong" feedback from the "agent is annoying" feedback, and address each one differently. Close the loop publicly - users need to see that their feedback changed something.
Heavy usage by one team only
Celebrate the enthusiast team as a reference, but don't rely on them for the renewal case. Pick one adjacent team with a similar job to be done, run a two-week guided rollout with the enthusiast team as peer coaches, and measure widening across at least three teams before the next QBR.
No exec championing
This is the most dangerous one because it's invisible in usage data. Re-engage the sponsor with a short, outcome-led briefing - one slide, one number, one story - and get a new commitment: either a named replacement champion or a steering cadence. If neither materialises, flag it to the Kainos account lead; this is a commercial risk, not a delivery one.
Linking adoption to credit ROI
Module 7 covered how Flex credits are consumed. This module joins the dots: adoption × credit cost = real value per credit. A credit spent on an invocation that the user overrides is a credit spent badly. A credit spent on an invocation that changes a decision, accelerates a cycle, or avoids a cost is a credit that earns its keep.
Credits consumed
From the Workday tenant - how many agent invocations, translated into the credits they cost.
Adoption quality
From leading metrics - what share of invocations were kept, not overridden, and tied to a real workflow.
Value per credit
Lagging outcome (time, cost, quality) divided by credits consumed - the number that turns the renewal conversation from cost to investment.
This is why we push customers off activation metrics and on to adoption metrics: a tenant with 100,000 invocations and a 40% override rate is spending credits that don't translate into outcomes. A tenant with 60,000 invocations and a 10% override rate is getting more value from fewer credits. Customer Success owns that story and should carry it into every forecast conversation with the customer's finance team.
Module 7 gave you credit forecasting. Module 10 gives you the denominator. The two together let Customer Success have an honest, numeric conversation about whether the customer is getting what they pay for.
How Customer Success teams integrate this
None of this works if it lives in a separate deck that only Customer Success reads. The cadence and the artefacts have to become part of how the account is run.
Cadence with customer
- Weekly operational stand-up for the first month - CS lead plus the customer's programme lead and agent owner.
- Monthly adoption review - CS lead, customer programme lead, a rotating business owner from a user team.
- Quarterly business review (QBR) - CS lead, Kainos account lead, customer exec sponsor, outcome-led narrative.
Artefact set
- Live adoption dashboard - leading metrics refreshed weekly, visible to customer and Kainos account team.
- Override log - sampled weekly in month one, monthly thereafter; themed into categories for intervention.
- Outcome tracker - one lagging metric per agent, tied to the original business case; updated monthly.
- QBR deck template - a fixed structure that pairs leading and lagging metrics with the next-agent conversation.
- Red-flag register - a short internal log of which red flags are live on which account, owned by the CS lead.
Customer Success holds the pen. Delivery supports. Account leadership uses the outputs to time the renewal and expansion conversation - not the other way around.
Check your understanding
Three questions. Each explains why every answer is right or wrong - the reasoning matters more than the score.
Next steps
- Pair this with Module 9 - Change management for AI. Adoption fails most often because the change wasn't managed; Module 9 gives you the prevention, Module 10 gives you the diagnosis.
- Revisit Module 8 - Adoption playbook. The detailed plays and artefacts Customer Success draws from sit there; this module is the frame that makes them purposeful.
- Connect to Module 7 - Predicting credit consumption. Adoption quality is the denominator in value-per-credit; neither number means much without the other.