We automate the processes your team still runs by hand.

AI agents read the request and its documents, check them against the systems you already run, and bring a team member in where a decision is needed. Nothing is replaced or migrated.

Free, with no commitment. A senior engineer replies within one working day.

Claude · Finance workspaceIllustrative example

Which customers have invoices more than 60 days overdue and an open support ticket?

Read open invoices in the ERP Matched accounts in the CRM Checked open tickets in the helpdesk

Three customers are more than 60 days overdue and have an open ticket. Together they owe $342,250.

CustomerOverdueOldestOpen ticket
Halvorsen Trading$184,20074 daysDelivery dispute 3 Nov
Meridian Freight$96,75068 daysPortal access 11 Nov
Castell Hotels$61,30063 daysInvoice query 14 Nov
Draft a reminder for eachTurn this into the Monday report
Reply to Claude

This is what one question costs your team today.

The data exists. It sits in three systems that do not talk to each other, so someone on your team becomes the connection.

  1. Export open invoices from the ERP to a CSV file.
  2. Export the account list from the CRM.
  3. Export this month's tickets from the helpdesk.
  4. Match the three files by customer name in Excel, fixing the names spelled differently in each system.
  5. Email the result. Repeat next Monday.
A desk covered in printed spreadsheets

Whole processes, from the first message to the finished job.

Claude reads the vague request, the scanned document and the supplier quote, checks them against your systems, and brings a team member in where a decision is needed. The result comes out in your template, checked by the right person, and seen only by those allowed to see it.

A laptop request that does not match the contract

A team lead asks for laptops in a chat message, without saying how many or which model. Normally nobody compares the supplier quote with the contract line by line.

Claude

Your documents, your decks, your voice

Out of the box, an AI assistant gives a correct answer in its own format, and someone on your team still has to rebuild it. We build in your templates, the wording legal has approved and the way each department reads its numbers. Switch between the two below.

Hartwell Marine1Finance committee · Week 46

Three accounts hold 342k of overdue receivables2

$342k360+ days overdue, each with an open support ticket
Halvorsen Trading184k
Meridian Freight97k
Castell Hotels61k
Source: ERP, CRM, helpdesk · 17 Nov44

    Each person sees what they are allowed to see

    The same question gives a different answer depending on who asks: everyone connects with their own account, so your existing permissions apply. How security works

    Which customers are 60+ days overdue with an open ticket?

    CustomerOverdueOpen ticket
    Halvorsen Trading$184,200Delivery dispute
    Meridian Freight$96,750Portal access
    Castell Hotels$61,300Invoice query

    Finance lead: all accounts, all amounts.

    Including the system nobody has a connector for

    Standard connectors do not reach a system on your private network, or the one your company built years ago. We build that connector. It runs on your infrastructure and the source code is yours.

    One we run in production: a connector for Microsoft Planner. A team member lists, creates and closes tasks from the chat, and sees only the plans their own Microsoft sign-in allows.

    How we build custom MCP connectors
    Jiraconnected
    SharePointconnected
    Teamsconnected
    Warehouse system (in-house, 2011)connector built by us

    It works inside the tools your teams already open every day

    Planning and delivery
    • Jira
    • Trello
    • Microsoft Planner
    • Asana
    • Confluence
    • GitHub
    Documents and files
    • Word
    • PowerPoint
    • Excel
    • PDF
    • SharePoint
    • Google Drive
    Conversation
    • Teams
    • Slack
    • Outlook
    • Gmail
    • WhatsApp Business
    Marketing and design
    • Canva
    • Figma
    • LinkedIn
    • Meta
    • your CMS
    Compliance and recordsCompany registries · screening providers · e-signature · your document vault
    Your own systemsDatabases, in-house applications and the tools nobody else has heard of

    What you can hire us for

    A process can run from a chat, from custom software we build for it, or as a custom workflow that starts on its own when an email or a form arrives.

    A project runs in phases: assessment, rollout, connectors for the systems that have none, and training and upkeep once it is live.

    We also build custom software and provide teams: web and SaaS, mobile apps, fractional CTO and dedicated teams.

    All services in detail, and the other work we take on

    How an engagement runs

    One project in three stages, at one price agreed before it starts. It begins with a free assessment, and what is built is yours.

    1. A wall of printed reports being reviewed by two people

      Assess

      We sit with the people who build the reports and run the approvals, map each one, and price the ones worth replacing.

      • Map of manual work
      • One quote for the project
    2. Many cables converging into one bright strand

      Connect

      AI agents for your team, wired to your systems, with your templates, your rules and your approval steps built in.

      • Connectors
      • House style
      • Approvals in Teams or Slack
    3. A team around a table in a bright office looking at a shared screen

      Run

      Training by department, permissions kept in order, and monthly upkeep, so it keeps working after we leave.

      • Training
      • Permissions
      • Monthly upkeep
    How an engagement runs, stage by stage

    Business process automation with AI: common questions

    What is business process automation with AI?

    Business process automation with AI means an AI agent carries a recurring process from start to finish: it reads the request and its documents, checks them against the company’s systems, does the routine steps and brings each decision to the team member who has to make it. Deep Blue Coding builds this on the systems a company already runs, without replacing them.

    Which business processes can be automated with AI?

    The best candidates are recurring processes where someone reads documents, looks things up in several systems and then asks for an approval: purchase requests, supplier invoices, client onboarding and KYC checks, weekly and monthly reports, and support follow-ups. A process is a good fit when its steps are slow because the case has to be carried from one tool to the next.

    How much does it cost to automate a business process?

    It depends on four things: how many systems the process touches, whether a connector has to be built for one of them, how many approval steps it has, and who maintains it afterwards. Deep Blue Coding gives one fixed quote for the complete project after a free assessment, before the company commits to anything.

    Can AI work with our ERP and CRM without replacing them?

    Yes. The AI agent reads and writes through connectors, so the ERP, the CRM and the helpdesk stay exactly as they are and nothing is migrated. Well-known tools already have connectors. For an in-house or legacy system, Deep Blue Coding builds a custom connector that runs on the company’s own infrastructure.

    Is it safe to connect AI to company systems and data?

    It is safe when access is set up correctly. The data stays in the company’s systems, each team member connects with their own account so existing permissions apply, and administrators decide which systems are connected and for whom. Anything that leaves the company, such as an order, a letter or a payment, waits for a named approval.

    Does AI automation replace the team’s decisions?

    No. The AI agent does the reading, the checking and the routine steps, and stops where a decision is needed. A team member approves or rejects in Teams or Slack with the reasoning attached, and the decision is recorded with name and time. A rejection sends the case back with the reason.

    Does AI automation only work through a chat interface?

    No. A chat is one way to start and follow a process, and it suits requests written in plain words. Deep Blue Coding also builds custom software and custom workflows: a web application with its own screens, a portal for clients or suppliers, or a flow that starts on its own when an email, a form or a new record arrives. The same connectors and approval steps sit behind all of them.

    What is a custom MCP connector?

    An MCP connector is the piece that lets an AI model read from and act on a business system through the Model Context Protocol, an open standard. A custom MCP connector is one built for a system that has no ready-made connector, such as an in-house application or a database behind a VPN. Deep Blue Coding builds them and hands over the source code.

    Who is Deep Blue Coding?

    Deep Blue Coding is an independent engineering consultancy that automates business processes with AI for mid-sized and large companies. It runs a process assessment, connects the AI to the company’s systems, builds custom connectors, custom software and custom workflows where they are needed, and trains each department. It works with Claude by default, and what it builds is not tied to one vendor.

    Start with one process.

    The report your team dreads building, or the approval that sits in an inbox for days. Tell us about it and a senior engineer tells you whether it can be automated and what it would take.

    Email social@deepbluecoding.com
    We reply within one working day.