--- name: Arco for humans version: 1.7.1 status: active owner: Stoirm Arnold last_updated: 30-Sep-2026 --- # Arco, for humans ## Built by humans, with AI. Arco started with a series of practical questions I was trying to answer in my own work: **How do I safely roll out AI without undermining the foundations of the business I’ve already built?** How do I stop useful knowledge, decisions and better ways of working disappearing into individual chats? How do I give AI enough understanding of the business to be useful without loading everything into every conversation? How do I preserve a good way of doing something so the next person — or the next AI conversation — does not have to work it out again? How do I let AI do more without losing control over what it can see, change or approve? How do I give people somewhere to develop thoughts and information privately before they become part of the organisation’s shared knowledge? How do I share what I learn about AI across the business without becoming the person everyone has to come to? How do I make AI work with the systems we already use rather than adding another layer of software? How do I avoid rebuilding everything if the best AI provider changes next year? And how do I build a system that notices when the business has changed, learns from real use and improves without someone constantly maintaining it? **Arco is the system I built to answer those questions.** It is an operating layer for humans and AI. It gives the AI you already choose to use — ChatGPT, Claude, or whatever comes next — the understanding, capabilities, boundaries and structure it needs to work properly inside your organisation. It is not another chatbot, task manager or replacement for the systems you already use. **It is what your AI runs on.** --- ## Who is Arco for? Arco is for leaders who want to use AI properly, but do not have endless time to work out how every part should fit together. It is for businesses that want to roll out AI **safely and effectively**, not just give everyone a licence and hope for the best. It is for organisations that have already tried AI and found the results disappointing: lots of experimentation, a few useful prompts, perhaps some clever demos — but no meaningful change in how the business actually operates. It is especially useful where there is no dedicated operations leader, systems thinker or AI transformation team whose job is to work all of this out. And it is for leaders who do have strong operators, but want the system to help preserve and apply the way they want the business to run without turning those people into permanent bottlenecks. The idea is simple: **AI should understand your organisation well enough to be genuinely useful, while still operating inside clear boundaries.** --- ## The problem I kept running into I work in operations. I spend a lot of time thinking about how businesses actually run: decisions, approvals, responsibilities, processes, context, handovers, follow-ups and all the boring-but-important things that stop organisations becoming chaotic. AI is brilliant at individual tasks. But a business is not one task. It is a collection of people, systems, decisions and rules that all interact. Without a shared operating layer, AI starts every conversation slightly blind. One chat knows something another does not. A process gets improved and then forgotten. Someone writes a great instruction, but it lives in one conversation. Two people ask the same organisational question and get different answers. A useful system quietly goes stale. That bothered me. I did not want ten new AI tools. I wanted **less systems, not more**. I wanted one chosen AI environment to understand how I operate and then keep that understanding useful over time. So I started building Arco. --- ## What Arco actually does Arco gives AI a maintained operating structure around the work it is doing. At a practical level, that means five things. **It helps the AI understand where it is operating.** Who are the people? How is the organisation structured? What matters right now? Which systems contain the authoritative information? That understanding can be maintained rather than reconstructed from scratch in every conversation. **It gives useful capabilities a repeatable structure.** If you work out a good way of doing something, Arco gives you a standard way to turn that into a maintained capability that can be found, used and improved again later. **It gives the AI clear boundaries.** Who can see something? Who can change it? Who needs to approve it? The answer should depend on the organisation using Arco rather than every AI interaction inventing its own rules. **It helps the AI find the right part of the system at the right time.** Arco should not load everything it knows into every conversation. It routes the AI towards the smallest useful set of context, capabilities and rules for the task in front of it. **And it gives the system a way to improve through use.** Corrections, failures, changes and better ways of working can become evidence for improving the maintained system rather than disappearing when a chat ends. Those are the jobs Arco is trying to perform. The individual Modules, Governance and Functions beneath it are the mechanisms used to do them. Arco does not try to prescribe every piece of work AI performs. It provides the structure through which useful AI capability can be defined, found, controlled and improved. --- # What Arco is made of Arco is built from separate parts, each solving a different problem, so the system can grow without becoming one enormous set of instructions. ## The modules ### Context — live Context gives AI a maintained understanding of your organisation, its people and the environment in which it is working. ### Skills — live Skills preserve good ways of doing recurring work so they can be used consistently rather than reinvented each time. ### Notes — early Notes give people a private place to capture useful information and develop it before anything important becomes part of the organisation's shared knowledge or actions. ### Personas — early Personas give AI a useful professional perspective when the quality of its judgement matters as much as the steps it follows. ### Decisions — draft Decisions will preserve what was decided, why it was decided and what later changed, so important organisational reasoning does not disappear into meetings and messages. --- ## Governance Governance sets the boundaries around what AI and people are allowed to see, change, approve or do. The aim is not to add bureaucracy. It is to give AI more useful autonomy without losing control. --- ## Functions Functions are tools Arco uses to keep the system reliable as it grows. **Dry Runs** test important changes before they reach the live system. **Defrag** finds and removes structural drift, duplication and outdated instructions. **Token Totter** checks that AI is not reading more information than it needs to do the job well. **Sitemapper** gives people and AI a simple map of where information lives. **Updater** brings improvements from Arco Core into an organisation without blindly overwriting what is specific to that business. **Daily Scan** reviews configured recent activity and routes useful findings into the parts of Arco you use. **Dev Scan** looks for evidence that the system itself needs to change. **Claudit** checks that Arco behaves reliably when used with Claude. --- ## Deployment and updates Arco separates the reusable system from the organisation using it. Each business has its own deployment, containing its local configuration and the parts of Arco it has chosen to use. That deployment is what its AI works from day to day. When Arco Core improves, those changes can be reconciled with the organisation's own deployment rather than simply replacing it. The destination is a system that can improve over time without erasing what makes each business different. --- ## Different AI providers Arco is designed to sit above the AI provider rather than depend on one. ChatGPT, Claude and whatever comes next may behave differently, but the organisation's knowledge, processes and controls should not have to be rebuilt every time the technology changes. --- # A system that improves Most business documentation gets less reliable with age. Arco is designed to notice when reality has changed — a correction, a failed process, a repeated problem or a better way of working — and turn that evidence into a controlled improvement. Ideas can be explored without immediately becoming rules, while old thinking can be retained without cluttering the system people use every day. The destination is a business operating system that becomes more useful through use rather than slowly going stale. --- # No new systems This is probably the shortest way I can explain the philosophy. I do not want Arco to become another platform people have to spend their day inside. Your documents can stay where your documents already live. Your finance system can remain your finance system. Your CRM can remain your CRM. Your people should continue doing their actual jobs. Arco should help the AI understand and work across that environment coherently. **Less systems. More useful AI.** --- # Why I am building it I built Arco because I wanted AI to work the way I think good operations should work. Clear ownership. Useful autonomy. Good judgement. Traceable decisions. Systems that improve instead of decaying. Enough governance to make things safe without creating bureaucracy for its own sake. Most importantly, I wanted to stop rebuilding the same understanding in every new chat. I want to be able to improve how I work once and have that improvement become part of the way the system works from then on. If that helps me run my own work better, it can help other leaders do the same. Arco is still early. Some parts are live and already useful. Some are being actively designed. Some ideas will almost certainly change as they meet real organisations. That is intentional. I would rather build the system from real use than try to imagine the perfect operating system in advance. ## Changelog | Date | Version | Change | Approved by | Changed by | |---|---|---|---|---| | 30-Sep-2026 | 1.7.1 | Included the promoted Daily Scan capability in the human explanation. | Stoirm Arnold — Defrag invocation | Codex Work | | 24-Sep-2026 | 1.7.0 | Replaced the opening question set so it reflects Arco’s current problems and roadmap, including durable knowledge, repeatable ways of working, Governance, private Notes, existing systems, provider independence and evidence-driven improvement. | Stoirm Arnold | ChatGPT Chat — GPT-5.6 Sol (Medium) | | 24-Sep-2026 | 1.6.0 | Reframed the human-facing architecture around the business problem each part solves; added Notes as an Early module; reduced Modules, Governance and Functions to concise destination-led descriptions; updated deployment, provider-independence and system-improvement language. | Stoirm Arnold | ChatGPT Chat — GPT-5.6 Sol (Medium) | | 21-Sep-2026 | 1.5.1 | Rolled the live changelog to the five-entry System Manual limit; older history remains preserved in the archived v1.5.0 release. | Stoirm Arnold | ChatGPT Chat — GPT-5.6 Sol (Medium) | | 21-Sep-2026 | 1.5.0 | Added Sitemapper to the human-facing Core Function map, describing its role in maintaining simple factual structural maps without turning mapping into documentation or automatic restructuring. | Stoirm Arnold | GPT-5.6 Sol |