Managed AI agent teams, run and verified by human experts
AI agents do the work. A named human lead owns the result. You buy an agreed outcome instead of headcount.
- Engagement
- Monthly
- Model
- Agents execute · humans stay accountable
- Teams
- Product BA Team · Product Development Team
- Oversight
- In the loop · on the loop · at the helm, chosen per process
- Pricing
- Base retainer plus a fee tied to the agreed result
- Status
- First pilots: two client teams onboarding by year end
- A team of agents with a named human lead accountable for outcomes
- Contracted around service-level objectives and a metric leadership agreed
- Weekly human review of real work; monthly report to leadership
- A chatbot subscription
- Offshore staff augmentation with an AI logo
- Agents with nobody accountable when they're wrong
- A black box you can't audit
Two teams launching now. A third on request.
Every team ships with a named human lead, an AgentOps layer and a metric agreed before day one.
Product BA Team
Product and operations leaders drowning in stakeholder requests
Agents do
Interview synthesis · requirements drafting · story and acceptance-criteria generation · backlog grooming
Humans own
Stakeholder validation · prioritization and risk calls · sign-off
Build-ready specs: synthesized interviews, requirements, user stories, acceptance criteria and a groomed backlog.
Product Development Team
Teams with a shipped product and a long roadmap
Agents do
Code generation · test writing · documentation · first-pass review · release notes
Humans own
Architecture · security · merge authority · production accountability
Working software released weekly, with tests, documentation and a full audit trail.
Custom Agent Team
Operations with a repeatable, measurable process
Agents do
Process execution inside explicit guardrails
Humans own
Exception handling · escalation decisions · quality verification
A team designed around one metric: vendor resolution, order exceptions, customer response time.
Run → Review → Tune → Prove
Agents execute. Humans review the real work weekly, tune what they find, and report the metric to leadership monthly.
- 01
Run
Agents execute inside explicit guardrails, goals and escalation triggers set by your human lead.
- 02
Review
Weekly human review of real conversations, specs and pull requests: the actual work, not samples.
- 03
Tune
Knowledge, guardrails and escalation rules updated from what the review found.
- 04
Prove
Monthly report on the agreed metric to leadership. The variable fee tracks it.
Review weekly · Report monthly
How much a human touches each decision
Every process the team runs gets an explicit oversight level. The riskier the call, the closer the human. The level is written into the contract and visible in the audit trail. It isn't a setting an agent can change.
- A named human lead signs off the level per process
- Escalation triggers route risky decisions to a person automatically
- Levels tighten or loosen only after a review, never silently
Human in the loop
The agent runs the process. A person owns every call that carries risk.
- Use when
- Customer commitments, money movement, regulated decisions.
- Human role
- Approves each consequential action before it lands.
Human on the loop
Humans supervise and manage by exception. Needs strong governance and oversight.
- Use when
- Mature processes with clear thresholds and a clean audit trail.
- Human role
- Watches dashboards, handles escalations, samples the rest.
Human at the helm
Humans set the agent's goals, limits and escalation triggers, then watch the audit trail for anomalies.
- Use when
- High-volume, low-risk work with tested guardrails.
- Human role
- Sets direction and limits. Steps in on anomalies, not on routine.
The performance and risk function behind every agent team
Observability, evaluation and drift detection are what make “human on the loop” honest. Without them, management by exception is management by hope.
Observability
Every agent action logged with inputs, outputs, cost and latency.
Evaluation sets
Golden test cases run on every change to knowledge or guardrails.
Drift detection
Quality and behavior tracked over time, with alerts when the agent starts to wander.
Escalation rules
Explicit triggers that route a decision to the human lead.
Audit trail
A reviewable record of decisions for leadership and compliance.
Cost governance
Token and tool spend tracked against the outcome it produced.
Priced on the result, not the hours
Base retainer
Covers the agent team, the accountable human lead and the AgentOps that keeps it honest.
Result-linked fee
Tracks the metric agreed up front, reported monthly with the evidence attached.
Contracts are structured around business outcomes, service-level objectives and performance metrics: shared accountability, not billed hours.
You get this
Every team is contracted backwards from these. If one isn't on the list, it isn't in the price, and we'll say so.
- 01 Results reported against the agreed metric
- 02 A named human lead accountable for them
- 03 A full audit trail of agent decisions
- 04 A monthly leadership report
Run connects to the other three
The Product BA team runs discovery: stakeholder noise to a signed hypothesis in days.
Explore ValidateThe Product Development team ships inside your squad, or takes over a live product to keep it improving.
Explore BuildEverything the team learns in production becomes material for your next cohort.
Explore Train“Two client teams are onboarding our first agent teams before the end of the year.”
What people ask about agent teams
Short answers to the questions that come up in the first conversation about this offering.
01 Who is accountable when an agent gets it wrong?
A named human lead on our side, written into the contract. That's the difference between a managed agent team and an AI tool subscription.
02 How is the fee tied to results?
A base retainer covers the team and AgentOps. A variable component tracks a metric we agree in advance and report monthly: cycle time, resolution rate, release cadence.
03 In the loop, on the loop, at the helm: what's the difference?
How much a human touches each decision. In the loop: a person owns every risky call. On the loop: humans supervise and manage by exception. At the helm: humans set goals, limits and triggers, then watch the audit trail. You choose per process.
Ask about a pilot
Thirty minutes to pick the process, the metric and the oversight level, and to tell you whether an agent team is the right move yet.