01OPERATIONS
The coordinator
Turn an agreed objective into a structured task plan, identify dependencies and prepare the next action for review.
Boundary: commitments stay with your team.
PERSONAL AGENTS / EXPANDED POSSIBILITY
Your next team
is an agent team.
Turn business intent into purposeful work. Explore AI agents that research, plan and assist—individually, or in a coordinated system with people firmly in control.
HIRING & CREATIONMEMORY & TOOLSLEARNING & EVALUATION
PASHARED CONTEXT
RESEARCHEvidence before answers.
PLANA goal. A bounded path.
REVIEWChallenge. Test. Refine.
ASSISTAct within permission.
A SYSTEM OF POSSIBILITIES · CONCEPT ILLUSTRATION
LESS REPETITION.MORE HUMAN AMBITION.
Hire a role. Create a capability. Orchestrate a team.
01 / THE AGENT ROSTER
Purpose first.
Personality optional.
Start with the job to be done. These role concepts help shape a brief—not a catalogue of instantly provisioned products. We qualify the tools, data and deployment before proposing an implementation.
02KNOWLEDGE
The researcher
Gather permitted sources, compare claims and assemble a traceable research brief with uncertainty left visible.
Boundary: evidence supports every claim.
03CUSTOMER
The service assistant
Classify incoming requests, retrieve approved guidance and draft helpful responses for a person to approve.
Boundary: sensitive cases escalate.
04OPERATIONS
The revenue companion
Prepare account summaries, organise meeting notes and suggest follow-up drafts from information you permit.
Boundary: no unapproved outreach.
05KNOWLEDGE
The evaluator
Test an agent’s proposed output against a defined rubric: accuracy, completeness, permissions and failure handling.
Boundary: a second model is not proof.
+YOUR BUSINESS
Your next agent
Bring a recurring task, a process bottleneck or a new idea. Build the brief around your people and the outcome.
Start small. Make usefulness measurable.
Create a mission brief ↗No matching role concepts. Try another search, or describe your own in the brief builder.
02 / THE ORCHESTRATION LAB
Many minds.
One accountable mission.
Explore how specialised roles can hand work to one another, share only the context they need, and stop at a human approval checkpoint.
MISSION CONTROLILLUSTRATIVE BROWSER DEMO / NO EXTERNAL ACTIONS
Compare permitted supplier information and produce a reviewable recommendation, with sources and open questions.
This deterministic sequence runs only in your browser. It does not call AI models, access your files or connect to business systems.
SHARED CONTEXTMission brief + approved source listScope stays attached to every handoff.
- 01DefineREADY
Agree the objective and permissions.
- 02ResearchWAITING
Collect permitted evidence.
- 03ReviewWAITING
Check the result against the brief.
- 04Human checkpointWAITING
A person decides whether to continue.
- 05Prepare the handoffWAITING
Package the approved draft.
Ready when you are. Start with a bounded mission.
03 / YOUR CONTEXT. YOUR CONTROL.
Choose where
intelligence works.
Deployment is a design decision. Explore the options to discuss with us; hosting, integrations and capabilities are confirmed during scoping.
DEPLOYMENT DISCUSSION / OC CLOUD
A shared operating layer.
A clearly defined boundary.
A managed-cloud deployment can bring coordination and approved tools together. Discuss tenancy, data residency, model providers, access controls, service availability and your operational requirements before selecting a design.
- Agree who can access which workspace and data.
- Define human review and external-action permissions.
- Qualify hosting, resilience and support arrangements.
A deployment concept to assess—not a claim that a cloud account has been provisioned.
UNIFIED MEMORY / DELIBERATELY SCOPED
Shared understanding.
Not unrestricted access.
Good memory is selective. Define what an agent may remember, where it came from, who may use it and when it should expire. Connect tools through explicit permission boundaries and keep consequential actions reviewable.
Remember with provenance
Keep a source, owner and freshness signal with useful context. Separate personal, team and client information.
Tools with a purpose
Start with the minimum access needed. A read operation and a write operation deserve different permissions.
People own the outcome
Define approvals for messages, payments, publication and other external commitments. Log the decision path.
Make forgetting possible
Plan retention, correction, revocation and deletion. Memory should be governed, not accumulated by default.
04 / THE AGENT UNIVERSITY
Better agents.
More capable humans.
An academy starts with practice. Use these introductory guides and evaluation habits to turn an interesting demo into a more disciplined experiment.
FOUNDATION / PRACTICAL GUIDE
Give an agent
a job worth doing.
Write a bounded task, an input contract and a useful definition of done.
Read the mission-design tutorial ↗PRACTICE / HANDOFF PATTERN
Design a team.
Not a group chat.
Specify handoffs, shared context and a clear stopping condition.
Read the orchestration tutorial ↗EVALUATION / TEST DESIGN
Test the failure.
Not just the highlight.
Build an evaluation set and a rubric before extending an agent’s authority.
Read the evaluation tutorial ↗BEFORE THE FIRST REAL TASK
Can your brief pass
these five questions?
A local self-check, not a certification or production-readiness assessment.
0 of 5 planning questions considered.
05 / THE MISSION BUILDER
Start with
one useful thing.
Use this local builder to frame your first conversation. It creates a text brief in this browser; it does not provision an agent or send an enquiry.
Nothing is sent until you complete and submit the enquiry form.
A LONGER HORIZON / BUILT THROUGH USEFUL WORK
The future of work
needs more than a model.
It needs understandable systems, useful tools, practical education and the confidence to know where automation should stop. We welcome conversations with businesses, builders, educators and investors who share that ambition.
06 / MAKE CONTACT
Bring the ambition.
Let’s shape the brief.
Tell us about the work, the people and the environment. We will use your enquiry to discuss suitability, implementation options and the capabilities available for your project.
↗
Enquiries are invitations to a conversation, not subscriptions or service orders. Please do not include passwords, confidential documents or sensitive personal data.
The demos on this site run locally in your browser. They do not connect to your organisation’s systems or execute real agent tasks.
CLEAR ANSWERS / NO MYSTIQUE
Understand
the boundaries.
What is a personal AI agent?
A personal AI agent is software designed to pursue a defined task using a model, instructions and permitted tools. Useful designs make scope, data access, review points and success criteria explicit. An agent is not a substitute for accountable human ownership.
Can agents run on my own machine?
Local execution is a deployment option to assess. Model size, available hardware, tool support, data policy and operational requirements determine what can run effectively. Cloud, client-managed and personal-machine routes must be qualified for your use case.
Is the orchestration demo connected to real agents?
No. The demo is a deterministic illustration running in your browser. It does not call a model, access business systems or execute external actions. It demonstrates a workflow with an explicit human approval checkpoint.
How do we start a project?
Choose one bounded business task, identify the data and tools it requires, and define a measurable outcome. Send a brief through the enquiry form to discuss suitability, implementation options and available capabilities. An enquiry is not a software subscription or service order.