The world of customer relationship management (CRM) systems and the general working world is currently dominated by a hot topic: AI agents. Commercials are popping up everywhere suggesting that one person can duplicate themselves into "umpteen clones" and thus revolutionize the way we work. But is this really a revolution comparable to electrification or the industrial revolution, or is it just temporary hype? In this podcast episode, Stefan Wendt and Dominik Enzler shed light on the current state of affairs, areas of application and challenges of AI agents.
In contrast to conventional chatbots or rule-based automation (such as workflows) that react to specific triggers ("If A, then B"), AI agents are characterized by their autonomy and ability to learn.
An AI agent:
Acts autonomously and recognizes tasks proactively.
Makes independent decisions based on the context.
Is capable of learning and adapts its actions and decisions based on previous interactions and experiences. Constantly expands its "neural network".
Understands company circumstances and can learn how the company talks to customers or how processes work.
The trend towards AI agents comes mainly from America. Major technology companies such as Microsoft, Google, Apple, Salesforce and HubSpot have jumped on the hype bandwagon and developed their own AI agents or co-pilots.
Microsoft Copilot: Well integrated into the Office world and can be used across the board, even if the depth of the learning function has been criticized.
Google Copilot (based on Gemini): Integrated into applications such as Gmail, suggests reply texts. However, the quality of the output is often perceived as inadequate.
Apple AI function: Offers email summaries and prioritization in Apple apps, for example.
HubSpot Agents: Directly integrated into the CRM and very helpful on an application basis. These include:
Customer Service Agent: Assists with support requests and uses knowledge databases.
Content Agent: Helps with the creation of landing pages, blog articles and research, accessing various LLMs such as Dolly, Midjourney, Perplexity or Claude. It is particularly valuable in the marketing sector.
Knowledge Base Agent: Searches company knowledge databases and suggests articles to reduce support requests.
Prospecting Agent: Supports outreach and company research, can generate individual e-mails. However, it currently only works in English, which limits its usability for the German-speaking market.
Social Agent: Helps with social media, schedules posts and suggests the best posting time.
Breeze: A general, built-in agent that makes suggestions for organizing everyday work.
Salesforce Agent Force: Similar to HubSpot agents, e.g. the SDR (Sales Development Representative) agent for company research and email creation.
Make (no-code platform): Allows you to create your own agents that can operate across different tools (e.g. Slack, ChatGPT, Gmail). This offers far-reaching use cases and great flexibility.
AI agents offer the potential to perform certain tasks faster and more efficiently. They can structure work better throughout the day and are particularly useful in areas that require deep research or enable personalized interactions based on unstructured data.
They are particularly strong in:
Marketing: the content agent can create blog articles and landing pages with appropriate images and generated content tailored to the context. This research work is very time-consuming and can be considerably simplified by AI.
First-level support: Since simple, already documented answers are often required here, a customer service agent (e.g. the HubSpot Service Agent, fed with company-specific documents and playbooks) can be very valuable.
Despite the great potential, there are considerable hurdles that make the widespread and unthinking use of AI agents difficult:
Monetization and costs: the development and operation of powerful Large Language Models (LLMs) are extremely computationally and energy intensive. The acquisition costs for servers are very high, and companies do not yet have a really good monetization strategy.
Credit systems: Many providers, such as HubSpot, work with credits. Every action of an AI agent consumes credits, which can quickly lead to exponential cost growth if usage is not carefully planned. A single blog article or a simple service response can cost several credits.
ROI calculation: It is crucial to individually assess where the use of an AI agent justifies the return on investment (ROI) and actually saves hours.
Language barriers: Many AI agents, such as the HubSpot Prospecting Agent, currently only work in English, which requires manual translation for the German-speaking market and hinders autonomous operation.
Quality of the output: The output of AI agents, such as Google Gemini, does not always reach the user's desired level of quality.
Data protection and compliance: A very critical issue, especially for German companies.
Data processing: When using AI agents, care should be taken to ensure that sensitive information is not sent unencrypted to the LLM providers or used to train the models. Providers such as HubSpot and Google are working to ensure data processing within Europe.
Be careful with your own agents: If you build agents completely yourself, information is often sent directly to providers such as OpenAI. It is advisable to deactivate the option for further model development.
Liability: A crucial point: The companies that use the AI agent are liable for incorrect or non-compliant information that the agent should share. Therefore, human verification of the results (e.g. for contracts) is essential.
AI agents are used in every area of CRM:
Sales: the prospecting agent can provide support, but sales remains a "people business". The interactions are often interpersonal and qualitative, which goes beyond quantifiable data. The use here must be carefully considered, as many sales activities would incur high costs.
Marketing: AI agents, especially the content agent, are much more interesting here due to their ability to conduct deep research and generate content (blog articles, landing pages).
Service: The customer service agent can help with support requests and search the knowledge database. This is particularly useful for first-level support and frequently asked questions.
To summarize: AI agents are not just hype, but offer revolutionary new possibilities for the CRM world and day-to-day business. They can make working methods more efficient and enable new products.
However, it is crucial to use and test AI agents carefully. Each company must individually assess which areas of application make sense in order to achieve a positive ROI and avoid "exponential cost growth". Issues such as data protection, compliance and the need for human control and liability must never be ignored.
CRM systems are getting much smarter. Now is the best time to "get a taste" and gain experience, because development will continue at a rapid pace.
Do you have questions about AI agents and their use in CRM? Then make an individual appointment.
If you want to stay up to date with the latest trends in CRM, subscribe to our free newsletter here.