Weekly quality measurement
We re-run your evaluation set every week and sample live conversations, documents or calls for human review. You see the agreed success metric, week by week.
Run
Go-live is the start. We stay.
A monthly retainer that keeps your AI systems accurate, safe and affordable after launch. We measure quality every week, fix what slips, adapt to model changes, and report to you in plain language.
- Typical length
- Monthly
- Indicative price
- ₹75,000 to ₹5 lakh per month, excluding GST
- Engagement path
- Discover→Prove→Build→Run
Sound familiar?
Typical situations, not quotes from clients.
- It worked well at launch. Three months later the team stopped trusting it, and nobody could say why.
- The model provider changed something and our answers got worse overnight.
- Our AI bill doubled and we didn’t know which feature caused it.
AI systems do not stay still. Customers ask new questions. Document formats change. Model providers update and retire models. Costs drift. Without someone measuring quality every week, accuracy decays quietly until people stop using the system.
What we do every month
Fixes and improvements
Failures from the weekly review become new test cases. Fixes are tested against the full evaluation set before release, so one fix does not break something else.
Model and vendor changes
When a model is updated or retired, we test the replacement on your evaluation set before switching. If a cheaper model scores as well, we tell you.
Cost control
Spend tracked by feature and by day, with alerts and limits. Pass-through costs reported line by line.
Monitoring and incidents
Uptime, latency, errors, guardrail triggers and unusual usage. Agreed response times for incidents, and a written review after any serious one.
Small changes, included
New intents, prompts, rules, document types or report formats, within the hours in your plan.
How it fits together
Where it fits
- Your cloud monitoring
- Your ticketing system
- Microsoft Teams or Slack alerts
- Your model provider accounts
- Your code repository and release process
We work in your environment and through your access controls. You can see everything we see.
How the engagement runs
Typical length: monthly. Minimum term of six months, then month to month with 60 days’ notice.
Discover
1 to 2 weeks, for systems we did not buildWe review the code, the prompts, the data flows and any existing tests. If there is no evaluation set, we build one first.
Prove
first monthA baseline report: current quality, cost and risk, and the top fixes.
Build
ongoingFixes and agreed improvements, released on a regular schedule.
Run
every weekThe weekly measurement cycle and a monthly review with your system owner.
Every system we take to production moves onto Managed AI by default. We also take on systems built by other teams or vendors.
How we measure it
We agree with you in writing, at the start of the retainer:
The success metric
from the original engagement, tracked every week against its target.
Safety checks
Zero-tolerance cases tested on every release.
Cost per transaction
Conversation, document, call or query, tracked monthly.
Response times
for incidents, by severity.
Changes delivered
against the monthly plan.
What you get
- Weekly quality report, one page
- Monthly review meeting and written summary
- A growing evaluation set, yours to keep
- Incident response within agreed times
- Model and vendor change testing
- Cost dashboard and monthly pass-through statement
- Release notes for every change
- A named lead at Envisstack
What we need from you
- A system owner who attends the monthly review
- Continued access to the environment, logs and model accounts
- Feedback from users, routed to one place
- Notice of upcoming changes to the systems the AI connects to
- Approval for releases, through your change process
Indicative price
₹75,000 to ₹5 lakh per month
excluding GST.
The fee depends on the number of systems, their volume, the review effort needed and the response times you choose. Minimum term six months, then month to month with 60 days’ notice. Billed monthly in advance.
Model, API and cloud costs are passed through at cost plus a 10% handling margin when we hold the accounts (no margin if you pay the providers directly), and reported separately every month. We never absorb them into the fee, so you always see what the system costs to run.
Frequently asked questions
Can you manage a system someone else built?
Yes. We start with a review and a baseline. If the system has no evaluation set, building one is the first piece of work.
Is Managed AI compulsory after a build?
It is our default, because systems that nobody measures decay. You can choose to run the system yourself. We will hand over everything and train your team.
What if quality drops below the agreed target?
We tell you in the weekly report, explain why, and fix it as a priority within the plan.
Can we move to outcome-based pricing?
Once we have several months of reliable measurement, we can discuss linking part of the fee to the success metric. Not before.
What happens if we end the retainer?
You keep the code, evaluation sets, reports and documentation. We hand over cleanly within the notice period.
Already have an AI system live? Let us measure it.
A 30-minute call with a founder. Tell us what is running and what worries you about it.