In short
- Start with the CRM’s own workflows for everything that follows clear rules: assignment, tasks, stage changes, notifications and sequences.
- Add AI where the input is free text or messy: classifying and summarising inbound emails, turning meeting notes into CRM fields, finding duplicates, drafting follow-ups.
- Salespeople stay in control: nothing is sent to a prospect or a customer without approval, and the CRM remains the single source of truth.
- Automated follow-ups must respect consent and marketing rules, and data protection applies to every record you enrich or store.
What native workflows do well
A workflow in HubSpot, a flow in Salesforce or an automation in Pipedrive is the right tool when a rule can describe the case. A form submission from a company with more than a set number of employees goes to a named owner. A deal that has not moved for two weeks creates a task. A closed-won deal triggers an onboarding checklist. These workflows are fast, cheap to run and easy to audit, and you do not need AI for them. We build them, and we review existing ones that have grown into a tangle.
Where AI adds value
Rules stop where the input is not structured. That is the daily reality of sales and service:
- Inbound emails and forms with free text: what does this person actually want, how urgent is it, which product line does it concern?
- Call and meeting notes: the outcome, the next step and the date are in the text, not in the fields.
- Duplicates and dirty data: two records for one company with slightly different names, or a contact with no company.
- Follow-ups and summaries: a short account of the history before a call, a draft for the next email.
Five typical workflows
These are typical processes, not client projects. In each, the AI prepares and a salesperson approves what leaves the building.
1. Lead intake
- Input: a web form, an email or a message from a prospect.
- AI prepares: reads the request, classifies the need and urgency, checks for an existing contact or company, merges obvious duplicates as a proposal, fills the fields and suggests the owner by your rules.
- Salesperson checks: the proposal for unclear cases and sends the first reply, which the AI drafts.
- Output: a clean record with the right owner and a reply draft.
2. Meeting notes to CRM fields
- Input: notes or a recording of the call, with everyone’s agreement.
- AI prepares: a summary, the outcome, the next step with a date, the people mentioned and updates to the deal fields.
- Salesperson checks: the changes before they are written.
- Output: the updated deal, a task for the next step and a note in the timeline.
3. Follow-up drafts after quotes
- Input: the quote, its status and the history with the contact.
- AI prepares: a follow-up draft after the interval you set that refers to the quote and the last conversation.
- Salesperson checks: and sends, or changes the plan.
- Output: a draft in the salesperson’s mailbox and a task.
4. Pipeline hygiene
- Input: the pipeline, with stale deals, missing fields and overdue tasks.
- AI prepares: a weekly digest per owner with proposed clean-ups and the reasoning.
- Salesperson checks: each proposal before any change is made.
- Output: a cleaner pipeline and a short digest.
5. Handover to delivery
- Input: a won deal with quote, contract and notes.
- AI prepares: the handover note and the customer record for the delivery team, with open questions listed.
- Account owner checks: the note and sends it on.
- Output: a handover pack in the system the delivery team uses.
What stays with people, and what we do not do
- Relationships and pricing: conversations, discounts and commitments are the salesperson’s.
- Outbound messages: nothing is sent to prospects or customers without approval, unless you have released a specific, plain message type.
- Scoring people for decisions about them: lead prioritisation with transparent rules is fine. We do not build systems that profile individuals in ways that affect their rights or access to services.
- Decisions about staff: we do not build systems that rate or select individual employees or applicants. These are high-risk under Annex III of the AI Act, see risk categories.
Consent, data protection and the AI Act
- Marketing messages: automated follow-ups to prospects must respect the rules on electronic marketing. In the UK, the Privacy and Electronic Communications Regulations apply. In Germany, § 7 UWG limits unsolicited advertising by email and phone. Other countries have similar rules.
- Personal data: every record is personal data. You need a lawful basis (Art. 6 GDPR), a data processing agreement with every provider (Art. 28), servers in the EU or UK and a way to honour the right to object to direct marketing (Art. 21).
- Enrichment: adding data from outside sources needs a lawful basis and transparency towards the people concerned. We use only sources you can justify.
- AI Act: Art. 4 asks for sufficient AI literacy, and Art. 50 requires chatbots to say that they are AI. This overview is general information, not legal advice.
Systems you probably use
CRM work usually happens in HubSpot, Salesforce, Pipedrive, Zoho or Microsoft Dynamics 365 Sales, connected to email, calendar, a quoting tool and the accounting system. We build inside the CRM’s own automation where it is enough and connect other tools through the CRM’s interfaces, automation platforms such as Make and Zapier or n8n, as described in n8n agency or DIY. The names are for orientation only and do not imply any partnership.
How to measure success
Look at a period before and after: the time between enquiry and first reply, the share of records with complete mandatory fields, the number of duplicates, deals without a next step and the time salespeople spend on admin. We agree the measures with you at the start and do not promise figures in advance.
Sources
- UK: Privacy and Electronic Communications (EC Directive) Regulations 2003
- Germany: § 7 UWG
- Regulation (EU) 2016/679 (GDPR), Art. 6, 21 and 28: EUR-Lex; Regulation (EU) 2024/1689 (AI Act), Art. 4, 50 and Annex III: EUR-Lex