4 Lessons On Selling AI to Enterprise Customers

Highlights from our AgentForce World Tour event in Paris.

Sam Ackah-Yensu, Nowi Kallen and Tassbieh Hassan

January 22, 2025

At our recent Agentforce World Tour event in Paris, industry leaders and enterprise practitioners gathered to discuss the practical realities of implementing AI in enterprise environments. Salesforce Ventures’ intimate roundtable session brought together Margaret Jennings, Head of Product at Mistral AI, Marvin Purtorab, Co-Founder & CEO of Convergence AI, and Mehdi Djabri, Co-Founder & CEO of Revo PM, alongside other enterprise technology leaders.

Here were four key lessons that emerged from our panel discussion:

1. Build Security and Privacy Into Your Foundation

Security requirements are crucial for AI startups selling to enterprise clients, particularly in Europe. Margaret highlighted how European regulations require PII classifiers to sanitize and anonymize data before processing — even code has to be treated as intellectual property.

The European regulatory environment stands in stark contrast to the U.S. market, where data privacy regulations are less stringent and data can more readily be used for model training.

This regulatory disparity creates both challenges and opportunities in the global AI market. Marvin emphasized that security concerns intensified when AI took direct actions. “When you have an agent that takes actions, people care much more about not just information security but output security — what happens if an agent breaks something or does something you didn’t want it to do?”

When serving customers in highly regulated regions or industries, successful vendors must prioritize:

Auditable AI systems
Clear data governance policies
Verifiable security controls
Flexible deployment models based on security needs:
- API-based access with strong trust frameworks and role-based access controls (a security method used to manage who can access certain information and resources based on their role within an organization)
- Cloud infrastructure through trusted providers
- Local deployment

For U.S. companies expanding globally, building security capabilities early provides a key competitive advantage, enabling the company to serve clients domestically and in more heavily regulated markets abroad.

2. Start Small And Scale Strategically

During our discussion, Marvin noted that a gradual approach to AI implementation helps identify valuable use cases: “Many people don’t actually have a concrete use case in mind. They want to use agents but need help discovering where agents can be valuable.”

Successful AI implementation requires methodical integration of data. As Mehdi explained, organizations need to add data sources “strategically, surgically, one by one.”

Revo’s approach to AI implementation entailed:

Starting with publicly available documentation
Gradually integrating core systems and systems of record like Salesforce CRM
Adding collaboration tools like Notion and SharePoint
Incorporating email and chat systems

Start with readily available data sources while building toward more complex integrations. As Mehdi emphasized, “The more you invest in data implementation, the more strategic your copilot becomes because the savings and empowerment you get from that is huge.”

Indeed, the effectiveness of any AI agent is linked to its data foundation — the more comprehensive and high-quality data it can access, the more intelligent and performant it becomes.

3. Focus on Differentiated Value Through Deep Customization

The key to standing out in the market lies in having AI that deeply understands your client’s specific workflows and processes.

Mehdi explained how organizations can achieve meaningful differentiation: “Right now, we’re not even tapping 10% of current LLMs’ potential. The potential comes from all the data you can connect and all the chains of prompting. It’s not about just doing one prompt, getting an answer, and doing a query. It’s about having multiple queries running in the background simultaneously on different sources and having them chained.”

Enterprise AI solutions that drive substantial results require organizations to:

Customize models through proprietary data and fine-tuning
Build prompts specific to the business
Connect to existing tools and workflows

Marvin demonstrated how customization drives real business value by sharing Convergence AI’s early work with Google’s HR department in the UK. By tailoring their AI solution to Google’s specific HR workflows and processes, the department achieved more than just efficiency gains. “We saw an overall increase in the number of tickets worked on,” he said. “People get to focus on the things they enjoy and offload the small things they don’t like.”

The urgency of customization became clear as Mehdi cautioned: “Those who don’t start training or exploring training their own copilot now will be left behind, because I’m seeing people becoming 10x performers thanks to solid copilots.”

4. Make ROI Clear and Measurable

While integrating AI requires upfront investment, market economics are becoming increasingly favorable. “The cost of inference, the cost of LLMs, has decreased dramatically,” Mehdi said. “Now it’s very cheap to execute high-value tasks at scale.”

Successful implementations balance quick wins that demonstrate immediate value with longer-term strategic advantages that can transform how teams work.

Beyond efficiency gains, Jennings emphasized a fundamental shift in work patterns: “You’re going from being an athlete to a coach,” highlighting how AI agents enables employees to focus on more strategic work.

The Implementation Era

We’re entering an era where AI can increasingly take meaningful actions on behalf of users. However, as Jennings emphasized, success requires building trust through reliable execution.

The future belongs to companies that can:

Build trust through consistent, reliable performance
Deploy AI that reflects their unique competitive advantages
Scale thoughtfully while maintaining security and data protection
Demonstrate clear business impact

Enterprise-focused startups should shift focus beyond pure technological capabilities. The winners won’t just be companies with the most advanced models, but those who can thoughtfully integrate AI to enhance their customers’ existing workflows while maintaining the security and trust enterprises demand.

Mehdi said that “By 2025, everyone will have an agent, and people will compete based on who has the best copilot.” The market advantage won’t be having AI — it will be having AI that truly understands and enhances your organization’s unique strengths.

Salesforce Ventures has a full schedule of 2025 events for founders and entrepreneurs to meet, network, and share knowledge. To learn more, sign up for our newsletter.

1. Build Security and Privacy Into Your Foundation

2. Start Small And Scale Strategically

3. Focus on Differentiated Value Through Deep Customization

4. Make ROI Clear and Measurable

The Implementation Era

Protex AI CEO Dan Hobbs On Thinking Big & De-Risking Entrepreneurship

Read Article

The New Enterprise Software Playbook: Lessons From Zeck and G2

Read Article