--- name: Arco how-to guide version: 1.0.0 status: active owner: Stoirm Arnold last_updated: 28-Sep-2026 --- # Arco How-To Guide Arco is designed to become more useful to **you** the more you use it. You should not need to learn filenames, understand its architecture or remember special commands. Most of the time, you simply tell your AI what you want in normal language. But there is one important thing to understand: ## Arco learns from your feedback Arco is constantly evolving. If it gets something wrong, tell it. If it keeps information that is not useful, tell it. If it misses something important, tell it. If its response is technically correct but not useful to you, tell it. The best feedback explains **what the problem is and why it matters**. For example: > You keep giving me every outstanding task when I ask for a handover. I only want the things that are still relevant today, because otherwise the useful things get buried. That is much more useful than: > Don't do that. Arco can use your feedback as evidence to improve the part of the system responsible for the behaviour. The more accurately you explain what works for you — and what does not — the stronger your Arco becomes. --- # Context **The problem:** AI often knows what you said in the current conversation, but not enough about you, your role or the organisation to make consistently good judgements. Arco's Context gives it maintained knowledge it can use again. That might include your responsibilities, the people you work with, how the business is structured, what a team does or how you prefer to work. Arco can hold both **shared company Context** and **local Context about you and how you work**. ### How this helps you You should not have to explain your job again every time you start a new chat. You can simply ask: > What do you know about my role? > Who is responsible for this? > Give me this advice in the context of what I'm currently working on. If something changes, tell Arco: > I've taken responsibility for finance approvals now. > Alex no longer reports to Sam. > This is important context for how I work — please update it. New or changed factual Context does not have to become accepted truth immediately. Arco can record the change and **submit it for verification**, so it can be checked against the appropriate source, evidence or person before it is treated as confirmed. That distinction matters: Arco should be able to learn quickly without quietly turning every statement into organisational fact. --- # Notes **The problem:** useful thoughts are scattered across conversations, meetings, voice notes and documents, while ordinary AI memory is not a reliable personal working system. Notes gives you a private working memory inside Arco. Use it for observations, thoughts, follow-ups, questions and things you are not ready to turn into formal facts or actions. ### Capture something quickly > Log a note: I think our transcript processing needs simplifying. > Note this for me — I'm not convinced the weekly meeting is still useful. You do not need to make it neat first. Arco can clean up the wording while preserving what you actually meant — including uncertainty. ### Your Notes build useful Context Your Notes do not have to become one enormous list. Tell Arco how you think about your work and it can organise Notes around that. You might want them grouped by: - client; - project; - person; - topic; - workstream; - or another structure that makes sense to you. For example: > Treat Client A as a client in my Notes and group anything relating to it underneath that. > Create a project view for the website rebuild. > Group my notes on AI adoption together as a topic. Arco can build those relationships without turning your private Notes into shared company information. That creates an important distinction: **Company Context** is shared information about the organisation. **Your local Context and Notes** help Arco understand how *you* work within that environment. ### Use it as a handover One particularly useful habit is a **daily handover**. At the start of the day, ask: > Give me my handover from yesterday. Arco can bring forward the useful unresolved threads from your Notes: things you were thinking about, following up, waiting on or intending to revisit. As those thoughts evolve, Arco should also clean up behind itself. If an older Note has been replaced by a newer understanding, it can be marked as superseded rather than continuing to appear as though both are current. If something has been completed, dropped or moved somewhere more appropriate, it can leave your normal working view while its history is still retained. This idea runs throughout Arco: **keep what is current easy to find, without losing the history of how you got there.** --- # Daily Scan **The problem:** useful information is constantly being produced around you, but you should not have to manually tell Arco what matters every day. Daily Scan gives Arco a recurring way to review recent activity and pull out useful changes, follow-ups, thoughts and information. A normal setup might include things such as: - your recent meeting transcripts; - email; - Teams, Slack or similar messages; - your Notes. But that is only the starting point. ### Make your Daily Scan yours The more useful question is: **What information do you regularly go looking for that Arco should be helping you find?** You might say: > Also check the sales call transcript folder and flag recurring customer objections. > Check my weekly board papers for anything that changes company Context. > Check this client folder for new meeting notes. > Every Friday, review the project status reports and identify things I've previously said are important. > Check my voice-note folder because that's where I tend to capture ideas. These can be baked into your Daily Scan rather than requiring you to remember to ask every time. ### Set it up > Set up my Daily Scan. Arco should help identify useful standard sources available to you and then ask what else would make the scan valuable. ### Run it > Run my Daily Scan. It should start from where the previous successful scan finished rather than repeatedly rereading your entire history. Daily Scan can feed the rest of your Arco. It may discover a new Note, identify a Context change, spot something that should be followed up or surface information relevant to one of your own modules. Finding something does not automatically make it true or public. A thought can remain a private Note; a factual change may be submitted for verification; an action may still require permission. --- # Skills **The problem:** most AI workflows slowly go stale. Someone writes a good prompt or set of instructions, it works for a while, reality changes — and the instructions stay exactly the same. Arco Skills are designed differently. Use a Skill normally: > Process these expenses. > Prepare my weekly update. > Run my client check-in process. After the work has been completed, Arco should consider whether the run exposed a **material improvement** to the Skill. Perhaps you corrected something. Perhaps a step was no longer needed. Perhaps Arco repeatedly had to ask for information it could have obtained itself. Perhaps the process worked differently in reality from the way the Skill described it. If something meaningful has been learned, Arco can update the maintained Skill so the improvement is available next time. If nothing meaningful changed, nothing needs updating. The result is a workflow that can improve through use rather than quietly ageing. ### Give feedback The same rule applies here: > This worked, but asking me for the client name was unnecessary because it's already in the source file. Update the process so it checks there first. Explain the problem and why the change improves the workflow. ### Adopt existing Skills You do not always need to build from scratch. Arco can provide reusable Skill templates that you can adopt into your own setup and then adapt through use. Ask: > Are there any existing Skills or templates I could use for this? Once adopted, your copy becomes yours. It can evolve around how you actually work rather than remaining permanently tied to the original template. --- # Personas **The problem:** sometimes you do not need another process. You need a different pair of eyes. Personas let you curate the expertise and perspectives you want available to you. They can be extremely personal and extremely specific. You might create: - a marketer who critically reviews your copy; - a cyber-security engineer who reviews a system setup; - a CFO who challenges the assumptions in a financial plan; - an employment lawyer who spots questions you should investigate; - an editor who understands exactly how you like to write. Invoke one naturally: > Get my Marketing Persona to review this. > Have the Cyber Security Engineer challenge this setup. You can also create a **panel of Personas**. For example: > Have my CFO, COO and Sales Director Personas independently review this plan, then compare where they disagree. That can be far more useful than asking one general-purpose AI for a generic critique. ### Make them yours Tell Arco what you like and dislike about their judgement. Give them examples. Narrow their remit. Remove behaviours that are not useful. A Persona should become a better representation of the perspective **you actually want available**, rather than a generic description of a job title. --- # Token Totter **The problem:** AI workflows tend to accumulate instructions. Every new exception gets added. Old guidance stays behind. Files become longer. The AI ends up reading large amounts of information for relatively simple tasks. That can make the workflow slower, more expensive and less reliable. Token Totter checks your AI workflows for that problem. For example: > Run Token Totter on my weekly reporting workflow. > Check these instruction files with Token Totter. > Why does this Skill need so much context every time it runs? Token Totter looks at what the AI actually needs to read to perform the task properly, identifies unnecessary weight and proposes the smallest safe improvement. It should not simply shorten things because they are long. A 20-page reference file that is only opened when needed may be perfectly efficient. A three-page instruction file loaded unnecessarily on every task may not be. --- # Dry Runs **The problem:** changing an AI workflow because an idea sounds better can easily make it worse. Dry Runs let you test the change first. Instead of immediately replacing your existing workflow, Arco can create a proposed new version and test it against the current one. For example: > I think this Skill would work better if it checked the CRM before asking me questions. Dry Run that change. Arco can: 1. preserve the current version; 2. create the proposed version; 3. run realistic tests against both; 4. compare the results; 5. identify whether the change genuinely improved the workflow; 6. only promote the new version once it has demonstrated that it is better. Your own Dry Run evidence should stay with your own Arco working area. This makes experimentation much safer. You can be ambitious with improvements without having to gamble the version that already works. --- # Sitemapper **The problem:** as your AI setup grows, both you and the AI can waste time simply working out where things live. Sitemapper creates a simple map of important locations. You might use it for your own project: > Build a site map for my client work. Or for a larger shared environment: > Update the site map so Arco knows where our finance, people and client information lives. The map helps Arco find things. It should not copy all of the information into another system. --- # Build your own modules This is where Arco can become much more than a set of company instructions. You can build parts of Arco specifically around the way **you** think and work. But the best place to start is usually **not by designing a system**. Start by doing the work. Get the result you want. Improve it. Then ask Arco to work backwards from what actually proved useful. For example: > This process worked really well. Turn it into something reusable. Arco should help decide whether what you have created is best represented as a **Skill, a module, a Persona or something else**. That is an important part of how Arco is built: **do the work first, prove the useful behaviour, then formalise the system around it.** Do not build architecture for problems you have not yet experienced. ### Imagine you wanted a Policies module Suppose you realised you repeatedly needed to know: - what policies currently exist; - who owns each one; - which version is current; - when it was last reviewed; - which policies are overdue; - what has changed over time. You could say: > I want Arco to help me manage our policies properly. Let's build this as a test module for me first. Arco can build the structure around the problem. It might create: - a simple register of policies; - stable identities so policies can be tracked reliably; - ownership and review dates; - current and historical status; - logs of meaningful changes; - rules for how new policies are added; - links back to the actual policy files wherever they already live. You could then ask: > What policies are overdue for review? > What changed in our Expenses Policy? > Who owns the Information Security Policy? Daily Scan could identify a new or changed policy. Governance could control who is allowed to approve or change one. The source policy files can remain exactly where they already are. **Arco builds the intelligent operating layer around them.** ### Or go much further **A Relationship Brain** Maintain important people, the last meaningful interaction, things you promised them, shared interests and when a relationship may need attention. **A Learning module** Capture useful ideas from books, articles and conversations, connect related thinking and surface relevant learning when you are actually working on something. **A Client Radar** Track changing priorities, unresolved issues, promises, important stakeholders and emerging risks across the clients you work with. **A Personal Decisions module** Keep the important decisions you make, why you made them and what happened afterwards — giving you evidence about your own judgement over time. **A Challenge Board** Maintain the big problems you are trying to solve and periodically bring together the right Notes, Context, Personas and evidence to challenge your current thinking. You do not need to know in advance exactly how any of these should work. Start with the real task. Use Arco to help you solve it. Then, once you know what works, ask Arco to turn that behaviour into something durable. --- # Why modules can work together Arco has been built in much the same way. Useful behaviour is proved first, then the common structure is worked backwards from what keeps appearing. That is why different modules share the same foundations. They have clear ownership. They keep a current register of what they contain. They preserve meaningful history rather than endlessly overwriting it. They keep logs where actions need to be traceable. They have clear rules for what can enter, change or leave. That common structure means one part of Arco can safely interact with another. Your Policies module could feed a colleague's compliance workflow. Your client information could interact with shared company Context. A team module could surface something into your private Notes without gaining access to everything else in them. The same underlying structure gives Arco a common language for connecting these systems while still respecting permissions and ownership. That is how Arco can become both a company operating layer **and your own Second Brain**. --- # Governance Some things should not be controlled simply because an AI — or a user — asks for them. Governance is the part of Arco that controls who can see, change, approve or perform things. Most users do not change Governance themselves. Instead, you experience it through sensible boundaries. Arco may tell you: > You can propose this change, but it needs approval before it becomes effective. or: > You can view this information, but you do not have permission to change it. The aim is not to add bureaucracy. It is what allows Arco to give you and your AI more useful freedom without quietly giving everyone unrestricted access to everything. --- # Decisions Some decisions are too important to disappear into a meeting transcript or an old email. Where Decisions is available, Arco can preserve what was decided, why it was decided and what later changed. You might ask: > Record this as a decision. > Why did we make this decision? > What replaced the old policy on this? Decisions is still developing, so the exact experience may differ depending on the Arco setup you are using. --- # What is happening behind the scenes? You do not need to manage most of Arco's internal machinery. Arco keeps current information easy to find while retaining older history. It uses registers so it can identify things reliably rather than repeatedly searching everything. It uses logs where actions need to be traceable. It keeps your personal material separate from shared company information. And it stores maintained information outside individual chats, so useful learning does not disappear simply because a conversation ends. The important part for you is the result: **your Arco can remember, adapt and improve without every useful lesson disappearing when a chat ends.** --- # The simplest way to get more from Arco Use it. Correct it. Tell it what matters to you. Explain why something did or did not work. Do the work first. Build the process second. Turn proven recurring work into Skills. Create Personas for perspectives you value. Use Notes as your working memory. Make Daily Scan look in the places you repeatedly go searching. Build small experimental modules around real problems you encounter. Use Dry Runs before replacing something that already works. And do not assume the system you have today is the system you have to keep. **Arco should get better as it learns how you work.**