Automation
An AI sales agent that works with your team and your CRM
An AI sales agent handles the work around each sales conversation, such as research, first drafts, follow-up reminders and CRM updates. Your salespeople keep the customer relationship and every decision.
What an AI sales agent actually does
At an office supplies distributor in Glasgow, five salespeople walk into the pipeline meeting. Many deals have no date for the last contact, because nobody logged the calls from the past few days, and the sales director spends the first half hour piecing together where each account stands. An AI sales agent takes on exactly that kind of work.
Technically, it is an automated workflow that connects a language model, the kind of program behind ChatGPT or Claude, to your sales tools, such as email, CRM, calendar and spreadsheets. It receives a task, looks up the data it needs, produces a result and puts it where your team will use it. That might be a completed company profile, a draft email or a call note filed against the right contact.
McKinsey's The State of AI 2025 found that 62% of organisations are experimenting with AI agents. In sales, the first useful cases tend to be modest: repetitive tasks with clear rules that someone does by hand when there is a spare moment, or skips.
Six sales tasks an AI agent can take on
These are the functions we design most often. Each company picks the ones that fit its sales process.
Research and enrich leads
When a new lead arrives, the agent gathers public information on the company, such as sector, approximate size, website and recent news. It fills in the CRM record and flags anything it could not confirm.
Draft first messages
It prepares a personalised draft from the record and your team's templates. The salesperson reviews it, edits it and decides whether to send it. Nothing goes out without approval.
Schedule follow-ups
If a prospect has not replied within the time you set, the agent creates a follow-up task and suggests a text. The salesperson gets the reminder with the context of the conversation.
Log calls and emails in the CRM
From emails and call notes or transcripts, the agent records the activity against the right contact. The customer history stays complete without copying and pasting.
Summarise each deal
Before a meeting or the weekly review, it writes a short summary with the last contact, next steps, objections raised and who makes the decision.
Flag stalled deals
It reviews the pipeline, meaning the list of open deals and their stage, and alerts the owner when one has gone too long without activity or a close date. The owner decides what to do.
Inside out: the order in which to automate sales with AI
Many projects start with the most eye-catching part, prospecting emails. We suggest the opposite order. Each ring opens once the one before it runs reliably.
- 01
First ring: what only your team sees
Logging calls and emails in the CRM and summaries before each meeting. If the agent gets something wrong, a colleague fixes it and the customer never knows. This ring also cleans up the data the next ones will rely on.
- 02
Second ring: what prepares the salesperson
Lead research, stalled-deal alerts and follow-up tasks. The agent proposes and the salesperson decides what to do with each alert.
- 03
Third ring: what reaches the customer
First messages and follow-ups written by the agent, always as drafts for the salesperson to review and send. This ring opens once the CRM is tidy and the team trusts the output of the earlier rings.
- 04
When to open the next ring
In each ring, count how much of what the agent produces is accepted unchanged. When that share is high and steady for several weeks, open the next ring. If it drops, adjust the instructions before going further.
What must stay with your salespeople
The customer relationship stays with your salespeople. They are the ones who listen, negotiate, adjust a proposal and know when to stop pushing.
That is why we agree clear limits with you from the start. A person always approves messages to customers and prospects. Your team decides prices, discounts and terms. Moving a deal to a new stage and the final qualification of a lead also stay in human hands.
The Harvard Business School and BCG study of 758 consultants makes the point. Those using AI produced work of over 40% higher quality and finished 25% faster. When they used it outside its strengths, results got worse. We put judgement about a particular customer in that category.
Risks of an AI sales agent and how we control them
A badly designed agent can harm your reputation faster than it helps. We deal with these five risks on every project.
Spam and mass sending
An agent can produce hundreds of messages in minutes. We set it up to prepare drafts, with volume limits and human review. Prospecting stays selective and your email domain keeps its reputation.
Data protection (GDPR)
We work with business data and public sources, keep only what is needed and honour opt-outs. The specific rules are checked with your data protection officer or legal adviser.
Brand voice
The agent writes from your style guide and from real messages your team considers good. Your salespeople's corrections are used to refine its instructions.
A messy CRM
If there are duplicates and stages each person uses differently, the agent spreads the mess faster. We review fields, duplicates and stages before automating anything.
An agent with no owner
Over time templates and offers change, and the instructions go stale. We agree with you on one person in the sales team who reviews the output and asks for adjustments.
How we design and build your AI sales agent
We build agents in n8n and connect them to the CRM and tools your team already uses.
- 01
Kick-off meeting and process map
We look with your salespeople at how a lead comes in, what happens to it and where time is lost. We pick one or two first-ring tasks to start with.
- 02
Baseline measurement
For one week, each salesperson notes the time spent on admin. We also record the response time to new leads and the share of CRM records that are complete.
- 03
Design with your team
We define which data the agent uses, what it produces, where it puts it and what a person must approve. Everything is written down before we build.
- 04
Build in n8n
We build the workflow and connect it to your email, your CRM and, if needed, your calendar or spreadsheets. We test it with real data in a controlled environment.
- 05
Pilot with a small group
Two or three salespeople use the agent for a few weeks. We collect their corrections and adjust the instructions and rules.
- 06
Documentation and handover
We hand over a documented workflow and train the person who will maintain it. Your team can keep improving it without depending on us.
What we measure after the pilot, and what drives the cost
After the pilot we measure the same things as in the baseline week and compare the two sets of figures.
We also track the agent's own indicators, such as the share of drafts sent with few changes, the stalled deals that were reviewed and the errors found in logged activity. If an indicator does not improve, we adjust that function or remove it.
The cost depends on the number of tasks automated, the state of your CRM, the tools to connect and the volume of AI model usage. We send you a quote and a timeline after the first meeting.
Frequently asked questions
What is an AI sales agent?
It is a system that connects a language model to your sales tools to carry out specific tasks. For example, researching a lead, drafting an email, logging a call in the CRM or flagging a stalled deal. It works within the rules you set. Decisions and contact with customers stay with your team.
Can an AI agent email customers on its own?
Technically it can, though we do not recommend it for prospecting or active customers. In our projects the agent prepares drafts and a salesperson reviews them before sending. That avoids mistakes, off-tone messages and mass sending that damages your domain's reputation. Unreviewed sends are limited to narrow cases, such as an appointment confirmation, and only with your agreement.
Is an AI sales agent GDPR compliant?
It can be, if it is designed carefully. We limit data to business information and public sources, keep only what is necessary, record where each piece of data came from and honour opt-out requests. The legal basis and retention periods are confirmed with your data protection officer or legal adviser. Our work does not replace that advice.
Which CRMs does it work with?
We build agents in n8n, which connects to many common CRMs and, through APIs, to email, spreadsheets and internal tools. In the first meeting we look at your team's CRM and check which data can be read and written. If we find limitations, we explain them before the project starts.
How much does an AI sales agent cost, and how long does it take?
It depends on the number of tasks, the tools to connect, the state of your data and the volume of model usage. The usual approach is to start with one or two contained tasks, run a pilot with a few salespeople and expand once the results justify it. After the first meeting we propose a specific quote and timeline.
Does AI for sales replace salespeople?
No. The agent handles preparation and record-keeping, such as research, drafts, activity logging and watching the pipeline. Conversations with customers, negotiation and sales decisions stay with your salespeople. Their time is simply split differently, with fewer hours on admin and more hours with customers.