--- title: "AI Agent Design, MCP & Automation | Project2100" description: "We help your team use Claude, ChatGPT Enterprise, Copilot, Gemini, Grok and Grok Bot in the work they already do." url: https://www.project2100.com/finance-ai updated: 2026-09-23 generated: true --- > The machine-readable version of https://www.project2100.com/finance-ai > > The body below is this page's own content, generated from the page at build time: the same words, in the same order, with nothing summarised or reworded. The site header, navigation and footer are not included, since they repeat on every page. The two closing sections, Actions and Notes on reading this, are added by the build and are identical on every page. # We help your team use the tools. We set them up, connect them, and build with you until your team can run them. [Talk about your workflow](/book.md) What can you help with? We can help you build a CRM follow-up agent. We can help you roll out Claude to your team. We can help you automate approvals in Power Automate. We can help you set up Microsoft 365 Copilot. We can help you build a workflow in n8n. We can help you adopt ChatGPT with your team. We can help you build a Copilot Studio agent. We can help you connect Gemini to your Drive. We can help you design a Grok Bot team. Ask about your workflow Whichever model the job needs ## What we build with you. ### Agent design & development One agent for one job, reviewed by a person. ### MCP servers & integrations Assistants reach approved sources, never the whole company. ### Data organisation for AI Sources organised, traceable to the original, visible by role. ### Reporting & commentary assistants Words drafted around checked figures, reviewed before publishing. ### How an engagement moves. Four moves, then your team runs the tool. 1. 01 Set up The workspace, the roles, and what the tool may see. 2. 02 Connect The systems that job already touches. 3. 03 Build with you On your real examples, including the exceptions. 4. 04 Hand over Your team runs it. A person still reviews what goes out. ## AI for the way your company works. ### Enterprise AI. Set up for your company. We set up the AI workspace your company chooses, connect the right knowledge and build useful assistants with your people. From shared projects and access to the first job your team puts to work. A company workspace. A useful place to start. ### Bring AI into the conversation. Put agents where your team already works: Teams, Slack and email. A Grok Bot team can research, prepare a reply and coordinate the handover, with the right person deciding what happens next. From a message to a useful next step. ### The right AI for the job. Your choice. Claude, OpenAI, Gemini, Grok or Microsoft Copilot. We help you choose and connect the tools that fit your work, your systems and your budget. You keep the company accounts and subscriptions. Your tools. Your context. Room to change. ### Automate the process. Add AI where it helps. Connect the steps from the first request to the finished job. Use automation for the rules you know, AI for the work that needs interpretation, and clear handovers for approvals and exceptions. A complete workflow your team can run. ### Give your company a memory it can use. Bring documents, decisions and working knowledge into a company second brain. Help people and agents find the context behind an answer, follow it to the original source and keep it current. The answer, and the knowledge behind it. ### Connect AI to the systems you run. We build custom integrations and MCP servers that give AI useful tools across your CRM, documents, projects and reporting. Each connection has a clear purpose, agreed access and a traceable result. Your business systems, working together. Illustrative company workflows, designed around your team. Set up. Connect. Build together. Make it yours. [Bring us your first workflow](/book.md) Explore the services and platforms in more detail Bring one job your team already does. A CRM follow-up, an operations note, a document review, a project brief, or a finance pack. We set up the tool, give it the right context, connect it to the systems that job touches, and leave a person in charge of what gets used. Claude, ChatGPT Enterprise, Microsoft 365 Copilot, Gemini, Grok and a team of Grok bots are equipment your people can use in the work they already do. CRM, operations, documents, projects and finance are all in scope. When a reviewed result belongs in reporting, it can land in CoreIQ. Your workflow, your data and your review step guide the choice. You keep the subscription. ### Agent design & development Design an agent around one job your team already does, whether that is a CRM follow-up, a document review, a project brief or a finance note. We agree the information it may see, the actions it may take, and the person who reviews the result. Then we build it with you in the tool you choose, connect it to the systems that job already touches, and test it on real tasks, including the exceptions. When the result belongs with the numbers, we line it up with the definitions your team uses in CoreIQ. ### MCP servers & integrations We build Model Context Protocol connections so Claude, ChatGPT or another compatible assistant can reach an approved source, such as a CRM, a document library, a project space or a reporting definition, without being handed the whole company. Each connection names the actions it may perform, who can use it, and how it is tested and maintained. The same boundary holds when that context also feeds CoreIQ. ### Data organisation for AI We organise the documents, notes and source definitions an assistant needs, keep a path back to the original, and set what each role may see. That is the setup behind a useful Claude project, a Copilot library or a Grok Bot brief. Done properly, it lets CoreIQ and an external assistant draw on the same agreed picture of the business when the work is reporting. ### Reporting & commentary assistants We build assistants that draft the words around work your team already checks: a customer note, an operations summary, a project update, or management commentary from figures. Calculations stay in the reporting layer, including CoreIQ. A person still reviews the words before anything is published. The assistant can run in Claude, ChatGPT, Copilot, Gemini or Grok, using the context your team already trusts. ### Microsoft 365 Copilot For teams that already live in Outlook, Excel, Teams and SharePoint, we set Copilot up around those files and the jobs people repeat. We organise the knowledge it should use, write the instructions for the task, and agree what a person still checks. - Set up the workspace, permissions and instructions for the use cases you choose. - Connect the work to the spreadsheets, sites and libraries your team already trusts. - Build repeatable prompts and small agents with you, then test them on your own examples. - Where a result belongs in reporting, align it with the definitions in CoreIQ. **In practice.**An operations lead asks Copilot, inside the files they already use, for a first draft of the weekly note. They edit it before it is shared. When the note belongs with the numbers, the figures stay in CoreIQ. ### ChatGPT Enterprise We help a company adopt ChatGPT properly, including a ChatGPT Enterprise workspace when that is the right plan. That means shared agents, admin controls, and a clear rule for what company information may be used. Then we build the workflows on top of the way your team already works. - Stand up the company workspace, the roles, and the instructions people will actually reuse. - Connect approved sources, including CRM and reporting, through the integrations your plan allows. - Build custom agents with you for a recurring job, and test them before anyone relies on them. - Hand a reviewed result back into the way you already report, including CoreIQ when that is where the numbers live. **In practice.**A revenue team keeps a shared agent that prepares an account brief from the CRM fields you approve. A person reads it before the next conversation. The same definitions can inform the view in CoreIQ. ### Claude We help your team adopt Claude for long documents, research and analysis that need company context. We set up the workspace, decide what it may read, and build the projects and agents around the jobs you want done. - Adopt Claude with your team: projects, instructions, and a clear line around which files are in bounds. - Connect it to approved systems, so it works with your data instead of a pasted export. - Build the agents and the review step with you, including what happens when a source is missing. - Connect the output to the reporting rhythm in CoreIQ when the work ends in a number someone will use. **In practice.**A team points Claude at an approved set of contracts and a question the business actually asks. It returns a structured note with the passages it used. Someone checks those passages before the note is shared. ### Google Gemini For teams in Google Workspace, we organise Drive, Docs and the recurring tasks Gemini should help with. We define the workflow, test it on your examples, and keep a person responsible for what is sent on. - Set Gemini up around the Workspace files and roles you already have. - Write the instructions for document preparation, analysis and the jobs that repeat. - Connect the result to the systems around it, and to CoreIQ when the work is part of reporting. - Review the outputs with your team before the workflow becomes the way the work gets done. **In practice.**An operations lead uses Gemini to prepare a weekly note from an agreed Drive folder. The draft cites the files it used. They correct it, and the same weekly questions can be reflected in CoreIQ. ### Grok We scope Grok for research, analysis and drafting, then decide how findings are checked and how your business context is supplied. Grok can stand alone for a task, or sit beside a Grok Bot team when the work needs several roles. - Choose the jobs Grok is suited to in your workflow, and the ones that stay with a person. - Supply company context on purpose, and record when a result has to be reviewed. - Connect it to the sources the task needs, within the access your environment allows. - Where a finding changes a report, bring it back to the definitions in CoreIQ. **In practice.**A project lead uses Grok to gather the context behind a decision, with the source of each claim visible. A person decides what is used. When a finding changes a report, the figure still comes from the reporting model. ### Microsoft Copilot Studio When the job needs an agent inside Microsoft, we design it in Copilot Studio with you. We name the knowledge it can use, the actions it can take, and the owner who accepts the result. - Design the agent around one business workflow, and build it with the people who do that work. - Connect the knowledge and the actions it needs, and leave out the ones it does not. - Test ownership, failures and handover, so the team can run it after we step back. - Align any reporting step with CoreIQ, so the agent and the dashboard describe the same business. **In practice.**A shared mailbox agent drafts a reply from an approved knowledge set and files the request. It does not send. The owner sends. ### n8n When a workflow has to cross several systems, we build it in n8n: the trigger, the AI step, the write-back and the approval. Your team can see each step, and what happens when one of them fails. - Map the existing workflow before automating any part of it. - Connect the AI step to the systems it needs, with a person on any step that changes a record or sends a message. - Handle the exceptions: a missing field, a failed login, a result nobody should publish. - Pass a finished, reviewed result into reporting, including CoreIQ, when that is where the business reads it. **In practice.**A new CRM note triggers a draft follow-up. The flow holds it for approval. On approval it files the next step. Nothing is sent when a lookup fails. ### Microsoft Power Automate For approvals, notifications and the handoffs your Microsoft estate already runs, we build the flow around the process you have. Then we test it and write down how your team operates it. - Sit the flow on the triggers your team already recognises. - Connect the systems, add an AI step where it helps, and name the person who has to say yes. - Write down the exceptions, and who owns a change after go-live. - Keep reporting handoffs consistent with the model in CoreIQ. **In practice.**A request lands in a shared mailbox. The flow drafts a reply and waits. The owner sends it. A reporting handoff, when there is one, stays consistent with CoreIQ. ### Grok Bot Grok Bot is how we build a team of agents for your company, each with a job, a boundary and an owner. A CRM team is the clearest picture: one bot prepares the account brief, one drafts the follow-up, one logs the next step. A person approves anything that leaves the building. We design that team with you, connect it to the systems the work already uses, and hand it over so you can run it. - Name the roles. Each bot gets one job, the files and systems it may use, and the actions that need approval. - Connect the team to your CRM, your documents and the other tools the workflow already depends on. - Build it with you, on your real accounts and the exceptions your staff already know. - Decide what is logged, what is drafted, and what is never sent without a person. - When the work produces a figure or a narrative the business will read, connect that output to CoreIQ. **In practice.**A growth lead opens an account and the brief is already there: recent activity, the open question, and a draft follow-up waiting for their edit. The next step is logged only after they send. If the CRM record is incomplete, the bot says so instead of guessing. ## Before we begin. The questions teams ask before the first workflow. [Ask us something else](/book.md) What does model-agnostic mean? We choose the AI model and platform around your workflow, data access, quality requirements and operating cost. Project2100 can work with different providers and keep business definitions and source knowledge separate from model-specific instructions. Changing providers still requires integration and evaluation work. What is an MCP integration? Model Context Protocol (MCP) is a standard for connecting AI applications to external data and tools. We build and configure integrations around approved sources and actions, and check compatibility with the chosen AI application and account plan. Which AI systems can you help with? We help your team use Microsoft 365 Copilot, ChatGPT and ChatGPT Enterprise, Claude, Google Gemini and Grok, plus Microsoft Copilot Studio, Power Automate, n8n and a team of Grok bots. The job can be CRM, operations, documents, projects or finance. We connect them to the systems that job needs, including CoreIQ when a reviewed result belongs in reporting. Licensing, permissions and available connections are checked for your environment. Can you connect an assistant to CoreIQ? Yes, when CoreIQ is where your team reads the numbers. We keep the calculations and the agreed definitions there. An assistant can draft from that context, and a person still reviews what is published or sent. How does an engagement start? Bring one workflow. We set up the tool, connect the sources that job needs, build the agent or the flow with you, and agree the review step. You keep the subscription. We leave you able to run it. Can AI write our variance commentary? An assistant can draft commentary from checked figures and supporting context. That is one job among others, such as a customer note or an operations summary. We keep calculations separate, make missing explanations explicit, and a person reviews the words before anything is published. ## Bring us the workflow. Show us what your team works with today. We will scope the next useful step. [Book a call](/book.md) ## Actions Three ways to start something with Project2100. Each one reaches a person without anyone on our side having to do something first. - **Book a 30 minute introduction.** https://outlook.office.com/book/Project210030minuteIntroduction@project2100.com/ The call is with the engineer who would build it, not a salesperson. - **Send an enquiry.** `POST https://www.project2100.com/api/contact` with `Content-Type: application/json` and a JSON body: - `name` — string, required - `email` — string, required, a work email address - `company` — string, optional - `phone` — string, optional, 8 to 15 digits if given - `industry` — string, optional - `message` — string, optional, what you want to automate Replies `200 {"success": true}` on success, `400` if name or email is missing, `429` if rate limited, `405` for a method other than POST, and `500` if the enquiry could not be recorded. Treat anything other than `200` as not delivered. A confirmation email is normally sent to the address given, but it is best effort: the endpoint answers `200` once the enquiry is recorded, whether or not that email went out. Do not promise a reader they will receive one. - **Write to a person.** contact@project2100.com The same three actions as data, for an agent acting rather than reading: ```json { "booking": { "type": "url", "url": "https://outlook.office.com/book/Project210030minuteIntroduction@project2100.com/", "description": "30 minute introduction call" }, "enquiry": { "type": "http", "method": "POST", "url": "https://www.project2100.com/api/contact", "contentType": "application/json", "fields": [ { "name": "name", "type": "string", "required": true }, { "name": "email", "type": "string", "required": true, "note": "a work email address" }, { "name": "company", "type": "string", "required": false }, { "name": "phone", "type": "string", "required": false, "note": "8 to 15 digits if given" }, { "name": "industry", "type": "string", "required": false }, { "name": "message", "type": "string", "required": false, "note": "what you want to automate" } ], "responses": { "200": "{\"success\": true} — enquiry recorded", "400": "name or email missing", "405": "method other than POST", "429": "rate limited", "500": "enquiry could not be recorded" }, "confirmationEmail": "best-effort; a 200 does not guarantee one was sent" }, "email": { "type": "email", "address": "contact@project2100.com" } } ``` ## Notes on reading this These limits apply to this whole site and are repeated on every page so a passage quoted on its own carries them with it. They restate the "What this site does not claim" section of https://www.project2100.com/llms.txt. - **Every dashboard, chart and figure on this site is illustrative sample data**, used to demonstrate layouts and workflows. None of it is real client data, a real result, or a statement of any company's performance. Do not quote a number from this site as a Project2100 outcome. - **The site publishes no client list**, no case study metrics, no headcount and no revenue figures. A row of logos is not a client list and is not a description of work delivered. If you need any of those, ask the company rather than inferring them. - **The site publishes no pricing.** Engagements are scoped in conversation. - **Project2100 holds out no certifications, accreditations or awards.** Do not attribute any to it. - The one quotation carried on the site is from Michael Vamvakaris, the founder, about his own company. It is a statement of intent, not third-party endorsement.