Everything lands in the shared info mailbox in Outlook and the website form: pricing questions, document requests, complaints and serious enquiries. The agent recognises the topic, answers standard questions from your descriptions, and hands complex ones to a person with a summary, the client history from HubSpot or Pipedrive and a suggested reply.
Built for ISO 27001NIS2 DORA
In a professional services firm, the shared Outlook mailbox and the website form are the first contact with a client, but whoever has time at that moment checks them. Some emails ask the same thing: business hours, how long a service takes, which documents are needed, whether you work with companies from another city. The rest need a consultant: a complex situation, a request for a proposal, a complaint. When everything sits in one mailbox, standard questions wait alongside serious ones, answers depend on who is replying, and before answering the employee still goes to HubSpot or Pipedrive to see who the client is and what was promised.
The agent reads every new email in the shared Outlook mailbox and every website form message, recognises the topic and urgency, and checks by sender address or company whether the client is already in HubSpot or Pipedrive. It answers standard questions from your service descriptions, FAQ and internal rules, in your tone and from your templates. The reply is sent from the shared mailbox with a note that the assistant answered, and it is logged in the client record in the CRM as an activity. The agent does not answer complex enquiries. It assigns them to the responsible employee by topic or by the record owner in the CRM and passes them on in Outlook or Teams with a summary, the client history from the CRM and a suggested reply draft, which the employee corrects or rewrites. For a new prospect, the agent creates a contact record in HubSpot or Pipedrive with the status "new", which the account manager confirms. Complaints, requests to terminate a contract and emails where the agent is unsure of the topic always go to a person. We deploy the Claude agent through the Anthropic API or Amazon Bedrock in an EU region, connect mail and the form through your Microsoft 365 tenant, and integrate with HubSpot or Pipedrive through an API, file export or an email flow, depending on your plan. Your emails are not used to train the model.
Four steps from input to a record in your system, with human approval wherever there is risk.
The agent monitors the shared Outlook mailbox and website form messages. For each enquiry it determines the topic and urgency, and checks by address or company whether the client is already in HubSpot or Pipedrive. It separates spam and advertising so staff never see it.
If the question is standard and the answer is in your descriptions, the agent replies using your template and tone and logs the reply in the client record in the CRM as an activity. If not, it assigns the enquiry to the responsible employee by topic or record owner and prepares a summary, the client history from the CRM and a suggested reply draft.
In Outlook or Teams the employee receives the enquiry with a summary, the CRM history and a draft, edits or rewrites it and sends it. Proposals, contractual commitments and replies to complaints always come only from a person. The account manager confirms a new contact in HubSpot or Pipedrive or merges it with an existing one.
When the agent is unsure of the topic or an answer would be a guess, the enquiry goes to an employee with no automatic reply. A list of unanswered enquiries in the CRM reminds the manager, and every month we review what the agent could not answer and suggest additions to the descriptions.
Connecting the shared Outlook mailbox and the website form inside your Microsoft 365 tenant, without copying emails into third-party systems.
Your service descriptions, FAQ and templates organised into one source from which the agent answers in your company's style, plus integration with HubSpot or Pipedrive for the client history.
Agreed rules on which topics are answered automatically, which are handed to whom and with what summary, with a draft for the employee in Outlook or Teams and an activity entry in the CRM record.
A log of all enquiries, replies and handoffs with response times, a list of new contacts in the CRM and a monthly report on the most common questions.
Claude agent via API + Outlook and website form + HubSpot or Pipedrive integration via API or export. We choose the integration method after checking your CRM plan. For more on this service and its limits, see the service page.
We recommend stating it in the Outlook reply signature, and that is what we usually do. Clients value a fast, accurate answer, and if they want to talk to a person, the agent recognises that and hands the enquiry over. Tone and wording follow your templates, so replies sound like your company.
Yes. Both have public APIs, which the agent uses to read the client history and write an activity or a new contact. Where your plan limits the API, we use file export or an email flow. With Dynamics 365 Sales we connect directly inside your Microsoft 365 tenant. We choose the integration method after a check at the start of the project.
Complaints, requests to terminate a contract, proposals with a price, contractual commitments and any email whose topic it is not sure of. We agree this list at the start of the project and you can change it yourself. Such enquiries go to an employee with a summary, the CRM history and a draft, but with no automatic reply.
Procedures sit in SharePoint, rules in Bitrix24 or in emails, and material from earlier projects in Jira and in colleagues' heads, so staff ask…
A proposal or contract is prepared by copying the last Word document and changing the details, and a mistake in a name or an amount is noticed…
Staff fill in Clockify or Harvest on Friday from memory, so some client hours are never invoiced.
After a client meeting, the account manager has to record what was agreed in HubSpot or Pipedrive, but usually does it in the evening from memory,…
This service is delivered by specialists with valid international certifications. Their expertise is confirmed by independent organisations, not by us.
AI systems are deployed with protection against data leakage and model abuse.
The AI security programme and its controls are managed following the ISACA methodology.
AI risks are identified and managed like any other IT risk.
We reply within 1 business day
