Five Steps for How GenAI Startups Can Break into the Enterprise: Observe.AI’s Playbook

The contact center is one of the areas in which GenAI will have an enormous impact. Stating this seems almost like a cliché, but how do we get from the early hype curve where we are today (see Klarna’s announcement) to an end state where the technology’s transformative power is realized? I’ve been fortunate enough to work alongside the Observe.AI team as they’ve laid out a blueprint for success in this space, having become the leading GenAI-powered company for the contact center. While the five-step process below has some specifics to its approach, it can also serve as a broader guide for other GenAI companies seeking to move into the enterprise.

1. Meet customers where they are

It’s easy to get wrapped up in the potential of GenAI’s once-in-a-lifetime wow factor. However, experimenting with customer experience with unproven chatbots is simply a non-starter for enterprises that have spent decades building their reputations. Since its inception, Observe has delivered an industry-leading QA platform, designed to ensure quality for agents. As customer service is able to automate away high percentages of rote requests, agent-led service for more complex issues will increasingly become a differentiator. By offering solutions such as AutoQA, text-based QA, as well as AI-driven, real-time agent guidance and coaching, Observe can meet customers where they are along their automation journey, rather than pushing an idealized version of the world that could get customers in trouble.

2. Integrate seamlessly with existing systems

Similarly, enterprises are not ready to immediately ditch their systems of record for shiny new technologies. Instead, they are increasingly willing to experiment with certain workflows, provided that they don’t create silos and still feed back into the main system. Ensuring that GenAI-powered workflows can work efficiently but still write back to the core system (e.g., ERP, CRM, etc.) is critical.

3. Architect for enterprise scale and performance

While LLMs are enormously powerful, they can also be ludicrously expensive and hard to scale. For some use cases, especially customer support, the latency within closed-source LLMs is unacceptable (think of how poor the experience would be of trying to resolve a customer service issue in real time with ChatGPT). If the end goal is an enormous deployment with a large customer, companies need to think ahead about how they’d actually be able to deliver at scale—as the potential buyer will work hard to validate this. For example, Observe is currently delivering one billion interactions per year across consumer brands like DoorDash, Public Storage, Cox Auto, and Omaha Steaks, which required enormous upfront investments in infrastructure. In addition, Observe’s custom LLM is trained on contact center data, which, at 40 billion parameters, is proven to significantly improve hallucinations and is 35% more accurate and faster in the contact center environment than the generic LLM models.

4. Transparency and trust win the day

Without data, there can be no advances in AI, and we’ve already started to see tensions in this area (see the New York Times lawsuit against OpenAI). This is even more important in the enterprise domain, where buyers need to ensure that their customer data and interactions are kept perfectly secure, and not used to train AI models that could be used by competitors. Expect additional scrutiny from customers, including additional buying cycles to go through AI ethics/safety counsels. Be willing to overinvest in data security and bend over backwards for customers (including deploying in a virtual private cloud).

5. Platform and partner

Everyone feels the strategic imperative for GenAI, from corporate boardrooms to technology partners who are determined to deliver exceptional experiences to their clients. Empowering those partners with GenAI is a surefire way to more rapidly gain trust with enterprises and take advantage of partner pull-through. Observe’s partnership with Concentrix allows them to deliver GenAI powered products to 125,000 agents across financial services, healthcare, retail, transportation and others in 15 countries.

While it’s critical to integrate into existing systems to get off the ground in the early days (see point 2), eventually enterprises want to know that you have the vision to deliver a true platform over time. They’ve jumped through enormous hoops and staked their reputations on an early vendor with new technology, and they eventually want to be rewarded for that. For Observe, having a full suite that includes everything from post-call QA to real-time AI gives companies like Block the confidence to invest in them as their AI partner for years to come as the technology continues to develop at lightning speed.

The potential for GenAI in the enterprise is enormous, but we are still in the early innings—at Menlo Ventures, we estimated in 2023 that GenAI spend was less than 1% of cloud spend. While enterprises know that generative AI will drive enormous efficiencies in their businesses, they remain cautious about adopting new vendors and providing them with key customer and business data. Observe’s success in moving to the enterprise provides a playbook for both startups and enterprises on how to leverage GenAI to win together.