a Private AI Practice
Commissioned agent systems for founders, family offices, and specialist professional teams.
a Private AI Practice
Commissioned agent systems for founders, family offices, and specialist professional teams.
What we are built around
Bespoke. By design.
Every agent is built around your context, your workflow, your decision logic, and your data. The agent is yours. It runs in your environment, on your data, under your control. The work is not a template; it is a fit.
You own what we build. All artefacts, configurations, prompts, integration code, and documentation are assigned to you on completion. Nothing we produce is held back.
The agent is fitted to you, not the other way around. The work begins with mapping how you actually decide, not with a pre-built template. The agent is the shape of your judgment, not the shape of our best practice.
Where we work best. Engagements where the decisions are high-leverage, the work is recurring, the information is confidential by nature, and the client is willing to invest a few hours early in the engagement to make the agent theirs.
Where we are not the right fit. Clients who want AI to make decisions unsupervised, who want a packaged tool without customisation, or who are evaluating AI as a hobby rather than as an operational change. We will say so in the first meeting.
Yours. By structure.
Your information stays where it belongs. These commitments are not features; they are structural to how we work.
Your data stays in your environment. We design agents to work in your cloud, your on-premise infrastructure, or a combination, depending on your data sovereignty requirements. Your data does not move to our systems, and we do not retain copies of your information beyond the engagement.
We do not train on your data. Your private information is not used to train any model — including for our own internal use. Where the agent itself is fine-tuned on your data, the training happens within your environment, on infrastructure you control, and the resulting model is yours.
Confidentiality is signed before the first meeting. All engagements are governed by a written confidentiality agreement signed before the Tier 1 meeting. You choose the agreement: we will sign your standard NDA, or use ours.
Subcontractors require your written approval. If a specialist contractor is needed, we name them and request approval before any work begins. We do not subcontract by default.
Independent. By commitment.
We do not take rebates, partner margins, or referral fees from any AI platform. Vendor-neutrality is structural, not aspirational.
Model-agnostic by design. We work with Anthropic Claude, OpenAI, open-source models, and others as the situation requires. The recommendation is what is right for the problem, not what is right for us.
The model selection is documented. When we recommend a model, the rationale is written down so the choice is auditable later. You can review and challenge the selection.
No platform reseller margins. Where we recommend a tool that you then procure, we receive no commission. The recommendation is the recommendation we would make if no vendor were paying anyone.
Vendor-neutrality is the foundation of trust. The work this practice does depends on the client trusting the recommendation. That trust is structural, and it is the practice's standard.
Visible. By practice.
The Mana System has been the operational reference for this practice since 2026. It is the system that runs the agency, and the system we use to design, build, and test every client engagement.
· Time recovered. A well-designed commissioned agent typically recovers 60–90 minutes per day of repetitive knowledge work for the principal user. Figures drawn from the internal Mana System deployment; not from a controlled client study. Client results will vary depending on scope, data, and adoption.
· Speed on knowledge-intensive tasks. Work that took 3–5 hours manually is typically completed in 30–60 minutes, including human review.
· Consistent output quality. The agent runs the same way every time, so the floor of quality is higher and the variance is lower.
· Autonomy within scope. The agent runs on its own within the boundaries you set. You do not have to be in the loop on every decision.
· A real artefact in your environment. The agent is yours, runs in your environment, on your data, with your governance.
About Mana Agency
Your Trusted AI Solutions Partner
Mana Agency designs, builds, and runs commissioned AI agent systems for clients whose decisions are consequential, whose information is confidential, and whose standards are non-negotiable.
What we deliver is an agent build, not a software licence — and the engagement is priced as one, not by the hour. Delivered as an agent that runs in your environment, on your data, under your control.
The work is shaped by a single belief: a token is not a decision. AI is good at generating options. People decide. The real value sits between the two — and that is where Mana Agency works.