--- title: Come il tuo team trae vero vantaggio da BusinessCards. | snapAddy url: https://snapaddy.com/it/resources/blog/ai-agents-in-crm lang: it description: Sai subito che numero di telefono ha la tua nuova collega? O a chi del team devi rivolgerti per ottenere il diretto di un referente? snapAddy BusinessCards rende tutto questo più semplice. E a breve aggiungerà una funzionalità che va ancora oltre. keywords: [agents, they, really, takeaways, glance] category: blog last_modified: 2026-07-30T07:56:10.354Z --- # AI agents in the CRM: When they really pay off. [Tutti gli articoli](https://snapaddy.com/it/resources/blog.md)Carla Diener il 30 luglio 2026 Two robots face each other against a light purple background, separated by a large question mark in the middle. A green checkmark under the left robot and a red X under the right robot illustrate the contrast between a preferred option and a non-recommended option. Hardly any topic is currently keeping CRM and sales teams as busy as AI agents. At our CRM Experience Conference this year, it was also one of the most discussed topics among the roughly 350 participants. That is no coincidence: if you do not address it now, you are not just missing a trend, you risk falling behind in the next wave of automation. CRM systems are meant to bring sales, customer operations, and campaigns together in one place. In reality, the work is still split across many separate processes: quotes, follow-up questions, approvals, customer requests. This is exactly where [AI agents](https://snapaddy.com/de/solutions/use-cases/ai-agents-for-sales.md) are now supposed to help and “think along” with you. But it is not quite that simple. Between the promise and the actual benefit lies a lot of clarification work, and that is exactly what pays off. In her talk “AI Agents in the CRM” at the [CRM Experience Conference](https://crm-experience.de/), [Liana Harutyunyan](https://www.linkedin.com/in/liana-harutyunyan-b973b069/) from the media group Axel Springer used four real-world insights to show what really matters. ## Key takeaways at a glance: - AI agents in the CRM are currently one of the most discussed topics in the industry, including at our own CRM conference. - 95% of all enterprise AI projects fail before they go live. The main reason is the process, not the technology ([MIT NANDA, State of AI in Business 2025](https://mlq.ai/media/quarterly_decks/v0.1_State_of_AI_in_Business_2025_Report.pdf)). - An AI agent can only automate what a company has already clearly understood itself. - Not every AI use case needs an agent; simple automation is often more than enough. - AI does not improve bad data, it makes it visible. **There is work between the promise and the benefit.** ## 95 percent of AI projects fail because of the process. According to [MIT NANDA](https://mlq.ai/media/quarterly_decks/v0.1_State_of_AI_in_Business_2025_Report.pdf), **95 percent** of all enterprise AI projects fail before they ever go live. This is rarely due to the AI itself. It is because nobody truly understood the process behind it before the technology was introduced. That is the most important lesson before you even think about agents: technology does not solve a problem that has not been clearly defined first. ## An agent only understands what you have understood yourself. Before a company builds an agent, it has to be clear what the real objective is. A simple tool for this is a **use case canvas**. At its core, it is a structured list of questions you answer before you touch any technology: - What exactly should work better in the end? - What does the current process look like, step by step? - Where exactly do effort and frustration arise? Only then do you get to the actual AI idea: which step can be taken over, and what data does the solution need to work with? This order is intentional. An agent that encounters an unclear process does not make it clearer; it just automates the confusion faster. ## Not every use case needs an agent. Many companies reach for an agent almost by reflex as soon as AI is involved. **Salesforce** has developed a simple evaluation model for this: the [Orchestration Density Framework](https://www.salesforce.com/blog/orchestration-density-framework-automation-decisions/). It answers, with three questions, whether you need an agent at all: - **Can you fully define the path to the goal in advance?** If yes, fixed rules are enough. - **Does the process require human judgment, or is a clear rule sufficient?** If it needs judgment, that points to an agent. - **Does the process work with clean tables and structured fields, or with free text and emails?** Unstructured content points to an agent. Your answers lead to a simple classification: Low means classical automation is sufficient. Medium means agent and automation work together. High means you need one or more agents that can make decisions independently. ## From fixed workflow to autonomous agent. The scale of low, medium, and high remains abstract until you apply it to a real case. Four examples from day-to-day CRM work show how these three levels differ in practice. 1. Automatically informing customers about a system outage is a fixed workflow with clear rules. **No agent needed.** 2. If a customer then replies with questions about the impact and possible credits, things change: you cannot predefine the response, it requires judgment. **Here, an agent is worthwhile.** 3. A pure price estimate based on fixed rules also does not need an agent. But if you first have to understand a customer’s requirements in order to build a tailored quote, you need the **interaction between agent and automation**. 4. At the top end of the scale is lead qualification: handling objections, scheduling appointments, running multi-stage email communication autonomously. Fixed rules are no longer enough here. **This is a clear case for an agent.** ## AI makes bad data visible, not better. Even a perfectly chosen use case will fail if the underlying data is wrong. An agent can only work with information that is actually available, up to date, and properly linked. This is where it quickly becomes clear whether the agent can deliver any value at all: an agent built on outdated or duplicate CRM records does not get better just because the technology is new. It simply exposes the true size of the data problem faster. ## A quick decision guide. Before you integrate the next AI agent into your CRM, three questions are worth asking: - Is the process truly understood and documented? - Does the Salesforce-based assessment actually show a high orchestration density? - And is the data foundation sound that the agent will operate on? If you answer these three questions honestly, you will not end up among the 95% of failed projects, but instead deploy agents exactly where they truly make a difference. At snapAddy we have developed a solution for this: **DataAgents**. Our workflow automation platform covers a wide range of use cases for sales and marketing and can represent any process in an individualized way. Human-in-the-loop ensures reliability and control. [Discover DataAgents](https://snapaddy.com/it/products/dataagents.md)