Implementing AI in Marketing and Sales: A Roadmap for B2B Teams

In short: The best way to implement AI for marketing and sales is in five stages: assessment, orientation, pilot, scaling, and return on investment. Clear processes, defined roles, and an AI-enabled tool stack determine whether initial tests will translate into measurable returns.

Where do B2B companies stand when it comes to AI in marketing and sales?

The majority have gotten started, but the potential remains largely untapped. According to Bitkom, 57 percent of German companies with 20 or more employees use AI. At the same time, 59 percent of users estimate that they are not tapping into the potential at all, while another 28 percent are tapping into it only to a limited extent (as of September 2026).

Marketing and communication, at 54 percent, are among the most common areas of application. In sales, 14 percent of companies use AI, up from 5 percent the previous year.

In everyday practice, it often looks like this: Individual colleagues chat with a chat assistant, the sales team tests another tool, and the CRM runs alongside it. Each person works with their own prompts, and the knowledge remains locked in chat histories. This creates pockets of productivity, but the sales pipeline sees little benefit from it.

The cost structure is revealing: 50 percent of AI users cite data preparation as a major cost factor, while 41 percent cite integration into existing systems. Twenty-one percent cite licenses for AI software. The effort, therefore, lies in data and processes. This is precisely where a structured implementation comes in.

What are the prerequisites for implementation?

Three fundamentals underpin every AI implementation: a well-maintained database in the CRM, clearly defined processes, and binding rules for handling data. You establish these fundamentals before onboarding the first agent.

We outline exactly what this means in HubSpot in the article “6 Requirements for AI with HubSpot Breeze.” To learn how to develop AI from an assistant into a growth driver, read “HubSpot AI: From a Useful Helper to a Real Growth Driver.”

How do you implement AI in marketing and sales step by step?

The LOTSE model breaks down the implementation into five stages: Situation Assessment, Orientation, Test Run, Scaling, and Return on Investment. The initial project runs for 60 to 90 days, with a fixed scope and documented results.

1. Situation Assessment: Where do you stand today?

The assessment maps out processes, data, potential, and cost structure in marketing and sales. You’ll see which tasks are time-consuming, which data is reliable, and where handoffs are problematic. This stage typically lasts two to four weeks. Your contribution: two workshops and access to your systems. The assessment belongs to you, along with all the results.

2. Orientation: Which use cases come first?

Based on the status report, we’ll develop a prioritized roadmap that you’ll decide on together with us. Each use case is evaluated based on its benefits, data availability, and implementation effort. Use cases with clear benefits and a solid data foundation will be prioritized. The follow-up costs of potential use cases are included in the results.

3. Test Run: How Do You Test Safely?

The following applies during the test run: one agent, one team, clear guidelines, and human approval. This allows a use case to prove its value under real-world conditions and within a manageable scope. Determine in advance which metric should improve and who will approve the results. This keeps the risk low and ensures the findings are reliable.

4. Scaling: How do you turn the test into everyday practice?

If the pilot proves successful, roll out the use case to additional teams and add new use cases. Now, training, documentation, and governance are key. According to Bitkom, 91 percent of companies using AI train their employees on how to work with AI. Enablement is therefore an integral part of any scaling effort.

5. Return on Investment: How Do the Benefits Show Up?

The business value becomes measurable in the results: time saved, better lead quality, shorter response times. These results are incorporated into the next status report. This way, your AI usage grows in cycles, each with proven benefits.

What roles are needed for an AI project in marketing and sales?

Four roles drive the implementation: a sponsor on the executive team, subject matter experts in marketing and sales, a person responsible for CRM and RevOps, and someone dedicated to enablement. In smaller teams, one person may take on multiple roles.

• Executive sponsor: This person sets goals, approves the budget, and decides on priorities. This gives AI the support needed for cross-functional change.
• Subject Matter Owners: Marketing and sales each designate a responsible person for each use case. This person defines requirements, reviews results, and approves agents.
• CRM and RevOps Responsibility: This role ensures data quality, integrations, and access rights. This ensures that all AI functions operate on the same, well-maintained data foundation.
• Enablement: This role trains teams, collects feedback, and maintains prompts and guidelines. This turns the tool into an active part of daily practice.

Acceptance within the team deserves special attention. According to Bitkom, 42 percent of AI users cite employee acceptance as a barrier. Early involvement and visible successes from the pilot phase build trust.

How can you tell if your tools are AI-ready?

Your tool stack is AI-ready when AI has access to complete, up-to-date customer data and feeds results directly back into your processes. Four criteria help with this assessment:

• Data quality: Contacts, companies, and deals are complete, up-to-date, and free of duplicates. Required fields and lifecycle phases are consistently defined.
• A single platform: Marketing, sales, and service all work within the same CRM. This allows AI to see the entire history of a customer relationship.
• Integrations: Email, calendar, ERP, and other source systems are connected. Data flows automatically between the systems.
• Rules and Access: You have defined which data AI is allowed to process and who uses which functions. Data protection and the AI Act are factored in from the very beginning.

Bitkom President Ralf Wintergerst describes the requirements for AI agents in similar terms: reliable data, connected systems, and clear rules for their actions. A platform like HubSpot consolidates CRM, marketing, sales, and service into a single database. The quickest way to check the status of your data is with a HubSpot Data Quality Audit.

Which use cases are best for getting started?

Use cases that are time-intensive, have clear rules, and yield measurable results are ideal for getting started. These examples meet the criteria in many B2B teams:

Marketing

• Content drafts: Blog posts, social media posts, and emails are created as drafts, which the team edits and approves. Measurable by the production time per asset.
• Data enrichment and segmentation: AI supplements company data and assigns contacts to your target audiences. Measurable by the completeness of the data records.
• Campaign analysis: AI summarizes results and highlights anomalies. Measurable by the time until the next decision.

Sales

• Meeting preparation: AI consolidates account history, recent contacts, and open deals into a briefing. Measured by the preparation time per appointment.
• Meeting Notes and Follow-ups: A summary and follow-up email are available as drafts immediately after the meeting. Measurable by response time.
• Prospecting with Approval: AI researches target companies and drafts the initial outreach; your team approves it. Measurable by the number of qualified appointments.

AI agents are taking center stage: According to Bitkom, 11 percent of surveyed companies are already using such agents, and 29 percent plan to do so. A test run with human approval is the ideal way to get started.

How do you measure the success of the AI implementation?

You measure success using business metrics: time saved, lead quality, response time, conversion rate, and close rate. Usage figures show adoption; business metrics show the return on investment.

Establish the baseline values in your status report and compare them after the pilot. Plan for at least one quarter to obtain reliable insights into the pipeline and revenue.

Questions and Answers About AI Implementation

How long does it take to implement AI in marketing and sales?

The initial project in the LOTSE model runs 60 to 90 days, with a fixed scope and documented results. The baseline assessment, as the first stage, typically takes two to four weeks.

What’s the best way to get started?

Start with the baseline assessment: processes, data, and cost structure. Afterward, select a use case for the pilot that is time-intensive, has clear rules, and yields measurable results.

Do we need HubSpot for the implementation?

The roadmap works regardless of the platform. A platform that consolidates marketing, sales, and service on a single data foundation significantly simplifies the implementation; HubSpot is one such example.

Start with your current state analysis

Implement AI where it delivers value. → Start the AI Readiness Check

Check the data foundation first? → HubSpot Data Quality Audit

Source: Bitkom e. V., press release “For the first time, the majority of companies are using AI” dated September 14, 2026, representative survey of 603 companies with 20 or more employees. View the press release