Skip to content
Engagement · Ongoing engineering support

Keep the system working after launch

Defined support coverage for a system Neul Labs built or hardened: maintenance, dependency and model upgrades, evaluation runs and reserved capacity for change.

Ask about this engagement How we work
The decision you are buying
Can you maintain the system and help us improve it?
Pricing
Monthly, quoted
Typical duration
Monthly, reviewed periodically
Afterwards
Support relationships often expand into adjacent workflows or further integrations — each scoped as its own decision rather than an assumed upsell.

What this engagement is for

AI systems drift in ways ordinary software does not. Model versions change, provider APIs are deprecated, prompts that worked last quarter degrade on new data and costs move with usage. A system that passed acceptance at launch needs someone watching the measures that made it acceptable.

Ongoing support is a defined arrangement, not an open-ended retainer. It names the systems covered, the response expectations, the evaluation runs that will be repeated, and the change capacity reserved each month. It only covers what the people and operating arrangements behind it can genuinely fulfil — if a commitment would need round-the-clock cover we do not have, we say so rather than sell it.

Support is usually offered for systems Neul Labs delivered or reviewed in depth. For other systems, a readiness review comes first so both sides understand what is being supported.

Right when

  • A system is in production and its owner wants continuity from the team that built it.
  • Evaluation and cost measures need to be re-run as models and providers change.
  • There is a steady stream of small improvements that do not justify individual sprints.

Not the right shape for

  • Systems nobody has inspected — those start with a review.
  • Mission-critical services that need guaranteed out-of-hours incident response beyond what is contracted.
  • Expecting support hours to absorb large new features, which are scoped separately.
Deliverables

What you receive

Maintenance

Dependency, SDK and model-version upgrades, applied and tested against the acceptance suite.

Evaluation runs

Scheduled re-runs of the evaluation set, with a short report when results move.

Reserved change capacity

An agreed amount of engineering time each month for prioritised improvements.

Monthly summary

What changed, what was measured, what is at risk and what is recommended next.

Process

How it runs

  1. 1

    Define coverage

    Agree systems in scope, response expectations, evaluation cadence and reserved capacity.

  2. 2

    Baseline

    Confirm the current measures so drift can be detected against a known starting point.

  3. 3

    Operate

    Apply upgrades, run evaluations and deliver prioritised changes within the reserved capacity.

  4. 4

    Review

    Review the arrangement periodically and adjust coverage as the system and its usage change.

What we need from you
  • The systems and components in scope.
  • The measures that define healthy operation — accuracy, failure rate, latency, cost.
  • Access to monitoring, repositories and deployment pipelines.
  • A named owner on your side for priorities and escalation.
How it is accepted
  • Coverage, response expectations and exclusions are written down.
  • Every upgrade passes the acceptance suite before release.
  • Evaluation results are reported against the agreed baseline.
Not included
  • Guaranteed out-of-hours incident response unless separately contracted.
  • New features beyond the reserved change capacity.
  • Systems or components outside the agreed scope.
Commercial terms
Support is quoted monthly from the systems covered, the evaluation cadence and the change capacity reserved. It is scoped to what can genuinely be delivered.

Situations that lead here

Launch production AI

The demo works. Now it has to launch.

Common follow-on →

Rescue an AI project

Stabilise an AI project that is losing confidence

Common follow-on →

Automate a business workflow

Make a demanding workflow shorter and dependable

Common follow-on →

Make AI useful for a team

AI tools are bought. Now make them useful.

Common follow-on →

Get an AI system approved

Give security something it can approve

Common follow-on →

Questions people ask

Do you offer 24/7 support?

Not by default. Support coverage is limited to what the roster and operating arrangements can genuinely fulfil, and the agreement states response expectations plainly.

Can you support a system you did not build?

Sometimes, after a readiness review. Supporting a system without understanding it would be a promise we could not keep.

What happens to unused change capacity?

The agreement states this explicitly. The usual arrangement is that capacity is reserved per month and does not accumulate indefinitely.

Other engagements: readiness or feasibility review · focused engineering sprint · embedded senior ai engineering. Compare them all on the engagements overview.

Can you maintain the system and help us improve it?

Monthly, quoted · Monthly, reviewed periodically

Ask about this engagement