The Agentic Revenue CompanyRevenue as a Service
Guide · Agentic sales

AI agents in sales

What AI agents in sales are, which tasks they take on, where people decide and how to introduce them successfully – with a focus on what matters: securing and growing revenue, with new and above all with existing customers.

What are AI agents in sales?

AI agents in sales are software systems that pursue a sales goal – such as securing repeat orders or reactivating dormant customers – and choose and carry out the steps themselves within approved limits. People set the goal and the scope of action and take the decisions that matter.

This is also called agentic sales. An agent needs four things: a goal oriented towards the business result, access to the data and tools the engagement provides, the ability to assess interim results, and an approved scope of action that it cannot extend on its own.

Chatbot, AI assistant, workflow, agent: the difference

Four forms of AI and automation in sales compared
What it doesWho steersTypical in sales
ChatbotAnswers questions in a dialogueThe person askingAnswering questions on the website
AI assistantSupports a person with a taskThe person, step by stepDrafting an email, summarising a call
Workflow automationRuns a defined processFixed rulesMoving a lead into the CRM, sending a reminder
AI agentPursues a goal and chooses the steps itselfGoal and approved scope of actionSpotting customers who are due, preparing outreach, matching replies and following up

The forms are not mutually exclusive. An agentic process often combines fixed rules, language models, interfaces and human decisions – and not every step needs a language model.

Which tasks AI agents take on in sales

The examples come from the documented state of our system. Which of them an engagement uses and which integrations are set up is defined for each mandate.

Recognising customers and demand
Agents analyse the data connected for the engagement and show where revenue can come from.
  • Calculating each customer’s purchase cycle from the order history in the ERP system and spotting customers who are due
  • Building customer classes from order frequency, interval and volume – approved by a person
  • Recognising dissolved companies from commercial register data and stopping outreach
Reaching out and following up
Agents prepare outreach, send according to defined rules and keep track of the next step.
  • First outreach from approved text modules that only mentions documented previous orders
  • Checking consent and the permitted channel before every message
  • Scheduling reminders and call tasks, and withdrawing them as soon as the customer responds
Understanding replies
Agents read incoming messages, match them to the right case and back up details with a quote.
  • Classifying replies in the inbox and extracting details
  • “No need at the moment” stops outreach without excluding the customer
  • Handing unclear cases to a person
Preparing and supporting conversations
Agents support salespeople before and during a conversation – they do not conduct it themselves.
  • Call briefing at the push of a button from cases, calls and emails
  • Opening the customer context for incoming calls
  • Live coaching during sales calls with a call guide from the approved playbook
Steering cases
Agents keep cases moving and suggest the next step. Decisions with external impact stay with people.
  • Daily case run with follow-ups and scheduling of the next case
  • Next best action from the playbook, with a reason
  • Sales figures on demand, such as revenue by territory compared with the previous year

All features with their scope →

How our agents work together

In the REVAGO system, several specialised agents work together. Two talk directly to your team; the others run in the background and pass tasks to one another. Anything with external impact or uncertainty goes through human approval.

PeopleAgents in dialogue with peopleAgents in the backgroundSystems and customersYour sales teamPeople decideApprovalsHuman-in-the-LoopCase Managerruns cases in chatSales Coachcoaches live on callsCustomer Analystdata, cycle, briefingMailer Agentsends after checksInbox Agentreads and matchesResearch & Knowledgeproposals from callsERP / CRMagreed systemsYour customersby email and phone

How the agents in the REVAGO system work together. Hover over or tap an agent to see its connections.

  • Dialogue with people
  • Task between agents
  • Path to approval
  • Approval and check point
Your sales team People decide
Your people in sales and inside sales. They talk directly to the Case Manager and the Sales Coach and take the decisions that matter.
Approvals Human-in-the-Loop
Human tasks decided under a personal login: unclear replies, suggestions from conversations and playbook drafts. Nothing takes effect until it has been approved here.
Case Manager runs cases in chat
Your team’s partner in chat. Finds and runs cases, plans the daily case run, chooses the next email – rules first, then the model –, commissions the first outreach and delegates analyses. Writes through the same paths as the user interface, never around the checks.
Sales Coach coaches live on calls
Supports your own side of the call live with a call guide and short prompts, then gives a review and practises in chat. Does not hear the other party and writes no cases or emails. Learnings go to Research & Knowledge as a proposal.
Customer Analyst data, cycle, briefing
Takes customers, contacts and orders from the agreed systems, calculates the purchase cycle by rule, prepares briefings from documented facts and checks whether a customer is still relevant – for example from commercial register data.
Mailer Agent sends after checks
Prepares drafts from approved templates and queues them. Before sending, the dispatch service checks consent, channel and address. Individual replies only go out once a person has approved them.
Inbox Agent reads and matches
Reads the inbox, matches replies to the right case, backs up details with a quote and processes delivery reports. Has no right to send; anything unclear goes to a person as a task.
Research & Knowledge proposals from calls
Analyses call transcripts and delivers only suggestions, follow-up questions and contradictions. Learnings from the Sales Coach become playbook drafts – effective only after human approval.
ERP / CRM agreed systems
The systems connected for the engagement. There is no universal integration; it is set up for each mandate.
Your customers by email and phone
Your customers only receive messages through checked paths and reply as usual – the reply reaches the Inbox Agent.

Schematic view of the sales roles that have been built, as of October 2026. Further roles for operation and development are not shown. Which roles an engagement uses is defined for each mandate.

Why existing customers are the strongest lever

Most guides to AI in sales start with customer acquisition: finding leads, automating cold outreach, booking meetings. But revenue also comes from customers you already know – when a customer reorders, commissions an additional service or returns after a pause.

Sales leaders see it similarly: according to a Gartner survey from May 2025, 57% count retaining and growing existing accounts among their top three priorities. At the same time, according to Bitkom (September 2025), only 5% of companies in Germany that use AI apply it in sales.

With existing customers, the order history and the relationship are already in place. That is exactly what agents need to recognise who is due again, which relationship is dormant and where additional potential lies. That is why we prioritise four use cases in this order:

Human-in-the-loop: where people decide

Autonomy and control are not opposites but designable parts of the same model. For every task, the right level is defined – depending on how large the impact is, whether a step can be reversed and how uncertain the situation is.

Prepare
The agent provides information or drafts, such as a call briefing.
Recommend
The agent suggests a next step with a reason; a person decides.
Execute with approval
A person approves, then the agent acts – for example with individual replies to customers.
Act autonomously within defined limits
For example editorially approved standard paths after rule, channel and consent checks.

More on human-in-the-loop in sales →

Introducing AI agents in sales

The introduction is less a technology project than a leadership project. These building blocks belong to it:

Clarify goal and mandate
Which area of revenue should be secured or grown, and who decides what? In AI-first coaching, you find out whether and how AI agents can secure and grow your revenue. If the answer is yes, you come away with a clear goal, a map and a mandate.
Define the use case and the result
One use case with a measurable result is a better start than ten ideas. What counts as a result, how it is measured and what the baseline is are settled beforehand.
Connect data and systems
Agents are only as good as the data they work with. Customers, contacts, orders and blocks come from the agreed systems, such as the ERP system. Data quality is checked before the first action.
Define rights of action
For every task, it is decided whether an agent prepares, recommends, executes with approval or acts autonomously within defined limits.
Start within a limited scope
The start takes place in an approved territory or customer group. This shows what works before the scope grows.
Run and improve
An agent is not finished once it goes live. Figures, feedback from salespeople and new cases feed continuously into rules, playbook and prompts.

Measuring costs and benefits

In a forecast from June 2025, Gartner expects more than 40% of agentic AI projects to be cancelled by the end of 2027 – because of costs, unclear business value or inadequate risk controls. The consequence: the value of an agent has to be defined before the start and measured in operation.

Open rates, call attempts and messages sent help to steer a process. But they are not a result. What gets measured is what reaches revenue – and what it costs. At REVAGO, in ongoing operation: you pay for results – not for activity.

Results as a Service: you pay for the result →

What AI agents in sales do not do

  • They do not replace a goal: if you pursue the wrong goal with AI, you only reach the wrong result faster.
  • They do not make up for poor data. Incomplete order histories or duplicates have to be cleaned up first.
  • They do not replace the relationship: people conduct conversations, negotiations and weighty decisions.
  • They are not a universal integration. The connection to ERP, CRM or inbox is set up for each engagement.
  • They do not work without operation. Rules, data and markets change – someone has to maintain the agent continuously.

How REVAGO implements AI agents in sales

We coach you. We build your agents. We run your system. In AI-first coaching, you find out whether and how AI agents can secure and grow your revenue. After that, we design the processes with your functional leads, build your agents, connect them to the agreed systems and run the system on an ongoing basis on our own software. The operational tasks carried out by people can sit entirely with you.

Frequently asked questions about AI agents in sales

What are AI agents in sales?

AI agents in sales are software systems that pursue a sales goal – such as securing repeat orders or reactivating dormant customers – and choose and carry out the steps themselves within approved limits. They analyse customer data, prepare outreach, match replies and plan the next step. People set the goal and the scope of action and take the decisions that matter.

What are examples of AI agents in sales?

Typical examples are an agent that recognises from the order history when a customer would be due to reorder and prepares the outreach; an agent that matches replies in the inbox to the right case; an agent that prepares a briefing on the customer and contact before a call; and an agent that suggests the next step from an approved playbook.

What kinds of AI agents for sales are there?

Broadly, there are three forms: agents as a feature within CRM or sales software, specialised tools for individual tasks such as outbound or booking meetings, and custom-built agents that are aligned with a business goal and your own systems and run on an ongoing basis. REVAGO belongs to the third form.

Can sales be replaced by AI?

No. AI agents take on tasks that can sensibly be automated – analysing data, preparing outreach, following up, matching replies. Relationships, conversations and weighty decisions stay with people. Designed well, salespeople gain time for exactly those tasks.

Which AI is best for sales?

That depends less on the model than on the goal. What matters is which area of revenue should be secured or grown, which data is available and who runs the agent. A tool is often enough for individual tasks; if you want a result, you need a system that is aligned with that result and improved continuously.

What is the difference between an AI agent and a chatbot?

A chatbot answers questions in a dialogue and waits for the next input. An AI agent is given a goal and chooses the steps towards it within defined limits – it reads data, plans cases and prepares messages, for example. In practice, an agentic process combines rules, language models, interfaces and human decisions.

How much do AI agents in sales cost?

Depending on the model, companies pay for licences, usage, project effort or results. At REVAGO, in ongoing operation: you pay for results – not for activity. No revenue or sales result, no fee. What counts as a result is set out in the contract; AI-first coaching is separate and can be booked on its own.

May AI agents contact customers on their own?

That is a design decision. At REVAGO, it is defined for every task whether an agent prepares, recommends, executes with approval or acts autonomously within limits. Only editorially approved standard paths run autonomously, after rule, channel and consent checks; individual replies only go out once a person has approved them. This does not replace a blanket legal compliance assurance – the legal review is part of every engagement.

Are AI agents also suitable for existing customers?

Yes, especially there. With existing customers, the order history and the relationship are already in place. Agents can recognise when a customer is due again, which relationship is dormant and where additional potential lies. That is why REVAGO prioritises repeat business, customer expansion and customer reactivation ahead of customer acquisition.

How do I get started with AI agents in sales?

With a goal, not a tool. First clarify which area of revenue should be secured or grown, and choose a use case with a measurable result. The first step at REVAGO is the short assessment: five questions as the basis for a conversation with a REVAGO coach.

Further reading