Home/Services/Mentat Autonomous Delivery

Services

Mentat Autonomous
Delivery

Short answer

Mentat Autonomous Delivery is a fixed monthly retainer for autonomous infrastructure work: pipelines, infrastructure as code, self-healing workflows, policy enforcement, and governed AI agents. Mentat’s AI-augmented engineers do the work on Mentat’s own stack, the inference and compute they use for that scoped work are included, and a human approves every promotion to production. You buy finished, production-grade outcomes and ongoing improvement, not tokens and not open-ended hours.

Most teams that try to put AI to work on infrastructure end up with a token meter they cannot predict and a backlog that does not move any faster. Stop feeding the token meter. Start feeding outcomes.

Autonomous Delivery turns that variable bill into fixed capacity. You agree the outcomes that matter, Mentat delivers them month after month inside deterministic guardrails (the model proposes, the guardrail decides), and a monthly outcome report shows what shipped.

Book an AssessmentTalk to Mentat

Why it exists

From a token bill to fixed capacity

Buying AI by the token rewards usage, not results. Buying engineering by the hour rewards time, not results. Autonomous Infrastructure Delivered is how Mentat describes its work, and this offering packages that work as a service you can budget for.

The inference and compute Mentat’s engineers consume while performing scoped work are included in the monthly fee. You do not manage model spend, prompt tuning, or agent tooling for that work. You get completed changes, reviewed and promoted, and a clear record of what moved.

  • A fixed monthly price instead of a metered one.
  • Completed, production-grade work instead of experiments and demos.
  • Continuous operational improvement instead of a one-time project that decays.
  • Mentat’s guardrails and human review in the path, instead of an agent with unsupervised access.

What is included

What the monthly retainer covers

  • AI-augmented engineering capacity: Mentat engineers working with Mentat’s internal stack to deliver and operate infrastructure changes.
  • Inference and compute Mentat’s engineers consume while performing scoped work, included in the fee.
  • Guardrailed agentic support for DevOps tasks, with every change proposed as a pull request your team can review.
  • Human-in-the-loop for production promotion: nothing is promoted to production without a person approving it.
  • A monthly outcome report covering what shipped, what changed in the environment, and what is next.

Inference and cloud costs

What is included, and what stays on your cloud bill

Included: the inference and compute Mentat’s engineers consume while performing scoped work for you. There is no token meter to watch for that work.

Not included: tokens, usage, and compute consumed by your own production systems, agents, or workloads running in your cloud accounts. Those stay on your cloud bill, because they run in your environment and under your control.

What the capacity works on

Work areas

The retainer is pointed at the outcomes agreed in the assessment. Typical work areas include:

  • Pipeline cleanup: consolidating, templating, and fixing the CI/CD pipelines that have accumulated.
  • Terraform and Ansible standardization: modules, patterns, and drift control so environments match what Git says.
  • Observability: signals, dashboards, and alerting that support self-healing workflows rather than noise.
  • Compliance and policy: policy as code that enforces rules in the path instead of in a wiki.
  • Azure DevOps health: access, service connections, agent pools, and classic-to-YAML migration.
  • Governed AI agents: agentic workflows that run inside deterministic guardrails, with an audit record of what actually shipped.
Azure DevOps Health & ModernizationPlatform EngineeringDevOps & DevSecOps

Guardrails

The model proposes. The guardrail decides.

Mentat builds the deterministic foundation first and lets probabilistic components operate inside it. Infrastructure as code, observability, identity, policy as code, and drift control come before more agents, because agents multiply whatever sits underneath them.

In practice that means a model can draft a change, but a rule layer decides whether the change is allowed to run, and a human decides whether it is promoted to production. If your foundations are not ready, the assessment will say so and the first months of capacity go to fixing them.

Artificial Intelligence at MentatThe Mentat platform

Pricing

Tiers and starting prices

Pricing is per month and fixed for the term. The ranges below are where engagements start. The final number is set in the assessment, once the first 90-day scope is locked.

Monitor & Evolve: starting at $5,000/month

For environments that mostly need to be watched, kept healthy, and improved a little at a time. Monitoring, health checks, and limited evolution of pipelines, IaC, and policy.

Core Delivery: starting at $7,500/month

Typically $7,500 to $10,000 per month. A steady stream of completed outcomes across pipelines, IaC standardization, observability, and policy, with a monthly outcome report.

Active Development: higher capacity

A higher-capacity tier for active development of new golden paths and governed agent workflows. Scope and price are defined in the assessment.

Minimum term

Three or six months. The assessment comes first, so the term starts with an agreed scope and a clear definition of done.

How it starts

Assessment first, then retainer

1. Paid discovery (the assessment)

A short, paid discovery that builds on the Mentat assessment: maturity score, ranked next moves, and a locked scope for the first 90 days, including outcomes, guardrails, and tier.

2. Choose a tier and term

The discovery moves straight into the retainer. Pick Monitor & Evolve, Core Delivery, or Active Development, and a three- or six-month minimum.

3. Monthly delivery

Mentat delivers against the agreed outcomes, with changes proposed as pull requests and checked by guardrails before anyone reviews them.

4. Review and promote

Your team and Mentat review each change. Production promotion is a human decision. A monthly outcome report closes the loop and sets next month’s priorities.

Ownership

You own what we build for you

You own the code, configurations, and evaluation sets delivered under the engagement. Mentat retains its internal playbooks and tooling, which are how the capacity is delivered. The aim is that you can run what we leave behind with or without Mentat.

Proof on the record

The track record this offering is built on

These are figures from Mentat’s published work across client engagements. They describe past results, not a promise that your environment will see the same numbers.

$30M+

Saved for clients

Savings delivered for Mentat clients, as shown on the Mentat homepage.

300k+

Automated pipeline runs annually

The kind of repeatable, code-driven delivery that governed agents can operate inside.

4 → 1

Outages reduced from 4/day to 1 per quarter

See the case studies for the programs behind these figures.

Related

Continue exploring Mentat

Cloud Automation AssessmentAll ServicesArtificial IntelligenceDevOps & DevSecOpsMentat PlatformAzure DevOps ModernizationPlatform EngineeringCase Studies

Frequently asked questions

Questions about Mentat Autonomous Delivery

What is Mentat Autonomous Delivery?

Mentat Autonomous Delivery is a fixed monthly retainer for autonomous infrastructure outcomes: pipelines, infrastructure as code, self-healing workflows, policy enforcement, and governed AI agents. Mentat’s AI-augmented engineers do the work using Mentat’s internal stack, and the inference and compute they use for scoped work are included. You buy completed, production-grade work and ongoing operational improvement, not tokens or open-ended hours.

How much does Mentat Autonomous Delivery cost?

Pricing is a fixed monthly fee. Monitor & Evolve, for monitoring and limited evolution, starts at about $5,000 per month. Core Delivery typically runs $7,500 to $10,000 per month. Active Development is a higher-capacity tier for new golden paths and agent workflows, and its scope and price are defined in the assessment. The final price is set once the first 90-day scope is locked, and the minimum term is three or six months.

Are inference and model costs included?

Partly, and the line is clear. Included: the inference and compute Mentat’s engineers consume while performing scoped work for you, so there is no token meter to manage for that work. Not included: tokens, usage, and compute consumed by your own production systems, agents, or workloads running in your cloud accounts. Those stay on your own cloud bill.

How does an engagement start, and what does the assessment cost?

It starts with the Mentat assessment, run as a short paid discovery that locks the first 90-day scope, outcomes, guardrails, and tier, and then moves directly into the retainer. You choose a three- or six-month minimum. Request it at https://www.mentat.cloud/assessment; Mentat confirms terms when you book.

Do AI agents change production on their own?

No. The model proposes and a deterministic guardrail decides what is allowed to run: policy as code, promotion gates, and an audit record of what shipped. Changes arrive as pull requests, and a human approves every promotion to production.

What kinds of work does the retainer cover?

Pipeline cleanup, Terraform and Ansible standardization, observability, compliance and policy as code, Azure DevOps health, and governed agentic workflows for DevOps tasks. The mix is agreed in the assessment and reviewed in the monthly outcome report.

Who owns the code and configurations?

You do. The client owns the code, configurations, and evaluation sets delivered under the engagement. Mentat retains its internal playbooks and tooling, which are how the capacity is delivered.

How is this different from hiring contractors or buying tokens?

Contractors are billed by the hour and token plans are billed by usage, so both reward activity. Autonomous Delivery is a fixed monthly capacity measured by completed outcomes and reported monthly. Mentat’s guardrails and human review are part of the service, and foundations such as infrastructure as code, observability, and policy come before more agents.

Start with the assessment

Replace the token meter with fixed capacity.

Book the assessment and we will scope the first 90 days around the outcomes you want. Prefer to talk first? Contact us and we will set up a call.

Book an AssessmentContact Mentat