A telecom operator in Belgium deployed a production AI sales agent in four hours. Not a proof-of-concept. Not a pilot. A system connected to their product catalog, internal databases, and WhatsApp — handling real customer conversations at scale.
The company is Orange Belgium, and the tool is Nexus, a Brussels-based startup that emerged from Y Combinator's Winter 2025 batch with $4.3 million in seed funding led by General Catalyst. What Nexus sells is a visual, no-code platform for building and deploying AI agents — and a library of more than 4,000 pre-built integrations with enterprise systems, from Salesforce to SAP to the messaging platforms where customers actually live. Non-technical teams are the target user.
A single Orange Belgium employee used Nexus to build and deploy a complete AI sales agent in one week, according to Nexus's case study page. The agent connected to Orange's entire product catalog and internal systems. It went live on WhatsApp, the messaging app most Belgian consumers use daily. Within four weeks it was running across multiple European markets. Orange reported conversion rates increased 50 percent, customer satisfaction scores climbed more than 10 points, and the agent now generates roughly $4 million per month in revenue — a figure Orange provided to Nexus and has not independently confirmed.
That deployment timeline is the point. Enterprise AI has a delivery problem. MIT research published in August 2025 found that 95 percent of enterprise AI projects fail to deliver measurable value — not because the technology doesn't work in the lab, but because the last mile from pilot to production is where projects go to die. Integration work, change management, organizational resistance, the months-long IT queue. A telecom employee shipping a live agent in four hours is not the norm. It is the exception — and Nexus is betting that exception is about to become the rule.
"We built Nexus to solve the last-mile problem in enterprise AI," said Assem Chammah, Nexus co-founder and CEO, in an interview. Chammah spent three years at McKinsey advising large organizations on AI strategy before co-founding the company with Shady Al Shoha, its chief technology officer. "The bottleneck was never the AI model. It was getting the AI to talk to everything else in the organization and getting a business team empowered to iterate on it without waiting for engineering."
That framing puts Nexus on one side of a fault line in enterprise software. Traditional enterprise AI implementations require consultants — often six-figure engagements lasting six months or more — whose revenue model depends on complexity being high and deployment being slow. Nexus, and platforms like it, are the counter-bid: pre-built connectors, visual workflow tools, and enough abstraction that a business analyst can ship what a consulting team used to build. This is not a new category — ServiceNow, Workato, and Microsoft Power Automate have been chasing this for years — but the agent layer adds genuine autonomy to workflows that were previously limited to simpler, rules-based automation.
Lambda.ai, a sales intelligence startup, is using Nexus to run agents across its sales and marketing functions, where it says a single agent saves hundreds to thousands of cumulative hours per month. The claim is imprecise, but the pattern is consistent: agents handling work that previously required human orchestration at every step.
The $4.3 million seed round is modest by enterprise software standards. General Catalyst led, with participation from Y Combinator, Transpose Platform, Twenty Two Ventures, and Phosphor Capital. Angel investors include Gokul Rajaram, Raphael Schaad, and Jake Mintz. The company operates from Brussels and San Francisco, and says it has more than 4,000 integrations with enterprise systems — a number it claims, unverified by third parties. Yuri Sagalov, Managing Director at General Catalyst, said in a press statement that Nexus is "enabling business teams to deploy AI agents without waiting on engineering," and that the firm sees this as a "fundamental shift in how enterprises automate workflows."
Whether it is fundamental or not depends on whether Orange Belgium is representative. The customer metrics — the $4 million monthly revenue figure, the 50 percent conversion lift — come from Orange, reported through Nexus's marketing. Independent verification is the gap: no public Orange Belgium statement corroborates these numbers, and the company's press office did not respond to a request for comment. The customer quote on Nexus's site is from Tom Guisgand, Senior Customer Journey Designer at Orange Belgium, who describes the experience positively but does not provide independent quantification.
The 4-hour claim is more durable — it appears on Nexus's own marketing site, which sets a higher bar for accuracy than an unattributed press release number. Even if the actual figure varies by a few hours in either direction, the order of magnitude is the claim: deployment in less than a business day, not a business quarter.
The systemic question is what happens to enterprise AI's delivery ecosystem if platforms like Nexus work at scale. Consulting firms whose engagement models depend on months of integration work are not the target customer. They are the displaced party. Whether that displacement is happening now or is still a few years out depends on whether the 95 percent failure rate is a technology problem or an incentive problem. Nexus thinks it is an incentive problem — and selling the solution to that problem is the entire business model.
