ThinkWork Enterprise · Services

Strategy, launch, and operations, for the Agent Harness.

We help teams scope, launch, operate, and expand governed agent workflows on the Agent Harness — whether you self-host or run it with us.

Scope a pilot See service packages

What you get with the harness

Production-grade Agent Harness

Threads, memory, agents, connectors, automations, and control — wired together, not assembled from parts.

AWS Bedrock AgentCore runtime

Sandboxed execution with Bedrock guardrails — not hand-rolled orchestration.

Per-tenant Cognito + IAM

Identity isolation enforced at the AWS layer, not in application glue.

Full audit + per-turn evaluations

Every turn traced; Reliability, Efficiency, Security, and Traceability evaluators run on the same trace.

One partner across the harness — open, operated, or services-led.

ThinkWork Enterprise is the services tier of the three-tier ladder. Whether you self-host the open Agent Harness on your AWS or run it with us as ThinkWork for Business, Enterprise wraps either model with strategy, launch, and ongoing operations — packaged as fixed-fee engagements, not billable hours. The shape of the engagement is named up front; the scope doesn't drift.

Common starting points

  • 01

    Support and triage

    Inbox, ticket routing, and first-response automation anchored in the workflows your team already runs.

  • 02

    Approvals and internal ops

    Multi-step approval chains and operational handoffs with human gates kept exactly where they matter.

  • 03

    Reporting and automation

    Scheduled reports, cross-system data stitching, and recurring workflows that replace manual pulls.

  • 04

    Connector-driven task flows

    Agents that act through Slack, GitHub, and Google Workspace — not just chat about them.

How an engagement looks

Audit. Build. Grow. Maintain.

We don't start by building random AI tools. We start by mapping where AI can produce measurable leverage — then build the highest-value workflows first, and keep them running.

  1. 01

    AI Audit

    A paid discovery and strategy engagement that inspects real workflows, scores opportunities against business value and implementation complexity, and produces a prioritized 90-day roadmap. Stakeholder interviews, workflow inventory, data readiness review, and a clear Phase 2 build proposal — decisions made, not a deck.

  2. 02

    AI Workflow Build

    The implementation phase. We turn the highest-value opportunities from the audit into working systems — internal assistants, operations automation, agentic workflows, or custom AI tools. Prototype → production build → pilot rollout → launch and handoff with full documentation.

  3. 03

    AI Development Retainer

    Ongoing AI strategy, development, and expansion for clients who want AI capability to keep growing after the initial build. We maintain your roadmap, ship new workflows on a cadence, optimize existing systems, and support adoption — month after month.

  4. 04

    AI Maintenance

    Lightweight ongoing support for teams with stable AI systems in production. We monitor health, handle small fixes and prompt updates, manage model and tooling changes, and send a monthly summary — so the system stays operational without a full retainer.

Services

Practical help from first workflow to ongoing operations.

Every package scoped up front — no billable hours, no open meter, no drift from the engagement shape we agreed to. Each one maps to a phase of the harness's adoption arc and names the components, controls, and operating guarantees it touches.

Fixed-fee Timeline · 2 weeks

AI Adoption Strategy Sprint

The first workflow, the first metric, the first rollout plan.

A two-week engagement that ends with three concrete artifacts — a named first workflow, a governance model expressed against the harness's five controls, and a 30/60/90 rollout plan. We pick the workflow that pays back fastest, agree on the templates that bound the agent's capability, and choose the success metric that proves it worked. No deck; the output is decisions you can hand to engineering on Monday.

  • Use case and workflow selection (with discard pile)
  • Templates + capability grants — Reliability + Security guarantees
  • Pilot success metric + the dashboard it lands on
  • 30/60/90 rollout plan with explicit gates
  • Risk register: what fails, how we'd notice, who handles it

Outcome

A named first workflow, a governance model expressed against the FiveControls, a rollout plan with concrete gates, and a one-page risk register — decisions made, not a deck.

Best for

Teams at the beginning who'd rather decide than discover.

Fixed-fee Timeline · 4–6 weeks

ThinkWork Pilot Launch

The first governed workflow, live in production.

Four to six weeks from kickoff to a governed workflow running in production. We stand up the deployment (in your AWS or on ThinkWork for Business), configure templates with model + guardrail pinning, wire the first connector, define the agent system prompt, build evaluations against real expected outcomes, and ship. By week four the agent is doing real work; by week six the audit, cost, and evaluation surfaces are tuned for steady-state operations.

  • Environment setup — open self-host or ThinkWork for Business
  • First agent: template + capability grants + system prompt
  • First connector wired (Slack / GitHub / Google / MCP / Email)
  • Evaluation suite seeded against real expected outcomes
  • Cost ledger + per-agent budget tuned to your guardrails
  • Launch + handoff with operator runbook

Outcome

One governed workflow live in production with its evaluation pass-rate tracked, its cost attributed per turn, and an operator runbook your team can drive without us.

Best for

Teams ready to move from planning to execution.

Ongoing operations Timeline · Ongoing

Managed ThinkWork Operations

Ongoing operations for a production ThinkWork deployment.

Recurring operations support that keeps the harness healthy, governed, and evaluated without you hiring a platform team on day one. Weekly we triage incidents, tune guardrails against real traffic, refresh evaluations as the workflow's expected outcomes evolve, and apply harness upgrades. Quarterly we review cost trends, spend efficiency, and evaluation drift against the original Reliability + Efficiency targets — and adjust the operating model when the data says we should.

  • Environment health and incident triage (weekly cadence)
  • Guardrail tuning + evaluation refresh — Reliability + Security guarantees
  • Harness upgrades + admin / configuration support
  • Per-agent cost reviews against budget — Efficiency guarantee
  • Quarterly operations review with explicit recalibration

Outcome

A production deployment that stays current, audited, evaluated, and operational — without a dedicated platform team and without ops drift between releases.

Best for

Teams running ThinkWork in production without dedicated platform ops.

Ongoing operations Timeline · Ongoing

Workflow Expansion Support

The next wave of workflows, shipped on a cadence.

Recurring delivery of new agents, workflows, connectors, and templates as the organization earns trust in earlier ones. Each cadence ships new agent templates calibrated against the same operating guarantees as the pilot, new connectors wired with the same credential-vault discipline, and new evaluation suites that prove each addition before it touches production traffic. The control model stays constant; the surface area grows beneath it.

  • New agent templates per cadence — capability grants + guardrails
  • Connector rollout (Slack / GitHub / Google / MCP / Email)
  • Template updates + skill-pack additions for fleet reuse
  • Evaluation suites for each new workflow before launch
  • Backlog prioritization + cross-team rollout coordination

Outcome

A steady cadence of new governed workflows shipping on top of your existing deployment — each one launched against evaluations, audited end-to-end, and traced back to the original control model.

Best for

Teams with early traction that want to keep shipping without reinventing governance per workflow.

Ongoing operations Timeline · Per plan tier

ThinkWork for Business

Same Agent Harness, operated by us — deployed in your AWS.

The middle door of the three-tier ladder: same harness, same controls, same six components — but you don't operate it. We run the runtime end-to-end inside your AWS account, take ownership of upgrades and incident response, and surface the same admin web your team would have run themselves. The product is identical to the open self-host path; only the operator changes.

  • Runtime operated by ThinkWork, deployed in your AWS
  • No AWS / Terraform setup on your side
  • Managed updates, upgrades, and incident response
  • Plan tiers that scale with usage
  • Same admin web, same audit log, same operating guarantees

Outcome

ThinkWork running as a managed product — deployed, updated, audited, and governed by us inside your AWS boundary, with the operator-facing surfaces unchanged from the open path.

Best for

Teams that want the harness, not the operating burden.

View plans

FAQ

Common questions.

How is this different from hiring an AI consultant?

A consultant delivers a recommendation. ThinkWork services deliver a governed workflow running in production — and optionally, the ongoing operating model that keeps it running.

What does the first engagement usually look like?

Either an AI Adoption Strategy Sprint (for teams still choosing where to start) or a ThinkWork Pilot Launch (for teams ready to ship a first workflow). Both are fixed-fee and time-boxed.

How does Enterprise relate to ThinkWork (open) and ThinkWork for Business?

Three doors into the same Agent Harness. ThinkWork is the open Agent Harness — Apache 2.0, self-hosted on your AWS, community-supported. ThinkWork for Business is the same harness operated by us — managed updates, priority support, still deployed inside your AWS. ThinkWork Enterprise (this page) wraps either with strategy, pilot launch, managed operations, and workflow expansion services. Teams move between tiers without rewriting workflows; only the operating model changes.

Do services cover both open self-host and ThinkWork for Business?

Yes. Enterprise services run on top of either deployment model. The harness, threads, memory, agents, connectors, automations, and controls are identical between them — only who operates the runtime differs.

Can we start small?

Yes — that's the default shape. Start with one workflow, prove value, expand from there. The governance model stays the same as surface area grows.

Do you only work on strategy?

No. Strategy, deployment, launch, governance, expansion, and ongoing operations are all in scope. The through-line is delivery, not advice.

Start with one workflow

One harness. One workflow. Running in production.

ThinkWork Enterprise is the services tier of the Agent Harness for Business. Whether you self-host the open Agent Harness, run it with us as ThinkWork for Business, or want both wrapped with strategy, launch, and operations — the starting point is an email. Tell us where you are and we'll come back with a shape.

Scope a pilot