Daniel Khachab on Choco’s AI Shift and the Competition for Talent

In October 2024, Choco co-founder and CEO Daniel Khachab discusses the food-distribution company’s shift toward AI with Harry Stebbings. His predictions about software, employment and Europe reflect his own views and his commercial interest in Choco.

Khachab says trying ChatGPT made him worry that AI would eventually replicate software built over years. Rather than wait for a competitor to disrupt Choco, he decided the company should disrupt itself and learn to build with AI throughout the organization.

He argues that the transition requires employees to acquire new skills, not only engineers. Stebbings counters that established relationships and proprietary data can remain advantages; Khachab agrees some defenses persist, but thinks the value of existing code may erode quickly.

Product design could change from constructing many screens to defining how an agent communicates and performs a task, he suggests. In his payroll example, a user might request a task in natural language rather than navigate a series of forms; this is a proposed future workflow.

He imagines selling software as an AI worker that can be scaled up or down, with consequences for sales and marketing as well as engineering. Such systems can respond differently to the same request, raising new quality-assurance challenges.

Khachab predicts that agent-based interfaces will eventually transform SaaS, though at different speeds across industries. A sales representative, for example, might provide notes and have a system update records rather than manually completing every field.

Stebbings questions whether traditional European enterprises will adopt this quickly. Khachab argues that conversational interfaces could reduce training barriers; for sensitive uses he suggests smaller or locally hosted models as possible options, not a universal solution to compliance.

Implementation and data readiness remain obstacles. Khachab says AI makes mistakes and lacks automatic knowledge of a company’s proprietary information, so products need a way for users to teach and correct the system. He cautions against promising perfection on day one.

On foundation models and applications, he describes Anthropic’s then-recent computer-use feature as a more capable form of automation. He expects competition among model providers to benefit application builders, while acknowledging that model providers may also build applications.

Choco uses mostly OpenAI, with some Anthropic and Mistral, he says. He estimates its model costs had fallen around 80% over six months despite higher transaction volume; this is a company-specific claim, not a universal price comparison.

The pair discuss automation and jobs. Khachab points to labor shortages in areas such as childcare and hospitality and hopes technology can free people for other work; Stebbings questions whether poor job conditions contribute to those shortages.

Khachab doubts companies can simply recreate and maintain all complex software internally with an LLM. Nevertheless he says Choco automated much of its customer care and account management and let hundreds of people go, including strong performers.

He describes those conversations as painful and says AI-related departures happened in one round. He presents the decision as a way to improve cash efficiency and pursue the company’s mission; that explanation does not remove the consequences for affected employees.

Khachab calls international competition for AI talent a “cold war.” He recounts invitations from government representatives seeking to attract AI teams and says some offers include visa help or, in extreme cases, salary support for several years; he does not identify the particular offers.

Choco began in Berlin because of its network and the city’s appeal to international employees. He says its R&D team is based there and estimates it has 15–18 machine-learning engineers. For application development, he believes teams can learn rapidly even without the same research institutions as London.

He worries that Europe sometimes receives AI products later and lacks sufficient chips, energy and competitive foundation models. These are his 2024 assessment and broad claims, not a complete inventory of Europe’s infrastructure or capabilities.

On chips, he argues for attracting leading manufacturers with government support, energy and water. On energy, he calls the transition difficult and expresses personal reservations about nuclear power while conceding it may be necessary.

He sees talent and long-term founder commitment as the hardest obstacles to competitive European foundation models. Stebbings points to European research institutions and Mistral as counterexamples; Khachab agrees success is possible but sees it as difficult to secure.

Khachab argues that European founders should not treat regulation or local funding constraints as automatic excuses for inaction. He says capital can be raised across borders, while acknowledging that regulation can impose genuine burdens, especially in certain industries.

He says the US is Choco’s largest revenue market and that the company recently opened an office in the UAE, receiving nonfinancial support such as visa assistance. He describes the US, Middle East and Europe as a portfolio of markets with different growth and implementation characteristics.

When COVID lockdowns began, Khachab recalls Choco had roughly $300–400 million in annualized gross merchandise value. He says 98% of GMV disappeared almost overnight as restaurants closed; that figure is his recollection of a company metric, not revenue.

He turned to crisis-management reading and sought to preserve humor and company values. Instead of hibernating and conducting mass layoffs, he says the team chose to keep pursuing business where restrictions allowed.

He recounts moving sales teams to locations including Miami and Texas, and adapting the go-to-market model from field sales to remote sales and then product-led growth. The company burned cash, but he believes continuing to operate helped it emerge ahead of rivals.

Khachab says the most recent round valued Choco at $1.1 billion and that taking the money helped the company through difficult periods. Stebbings counters that excess capital can create distraction and reduce discipline; Khachab accepts those risks.

His biggest regret about unicorn status is cultural: employees or applicants may assume the company has already made it. He recalls holding a unicorn party and later removing the decorations because he wanted the team to remain hungry and willing to take risks.

Looking back at the first week of COVID, he regrets giving everyone a week off without a clear purpose. The pause helped management regroup, he says, but he believes the team needed direction during the uncertainty.

On work-life balance, Khachab prefers full attention to the work at hand and to home when away. He and Stebbings also acknowledge that urgent business demands can interrupt personal time; he sees presence as important to avoiding burnout.

He says Choco has raised $330 million. His fundraising advice is to build a strong team, mission and operating results first, then present the evidence clearly; founders should approach investors as partners rather than beg for money.

Khachab believes a compelling long-term mission helps a founder lead through hard periods. He calls for founders willing to devote 15–20 years to building enduring businesses, a demanding personal philosophy rather than a rule for every company.

Asked whether he would start again, he says his answer has varied with the year. He advises would-be founders to be honest about how difficult the journey can be and whether they feel strongly enough about the work to persist.

As a former ski instructor, he remembers that children willing to fall repeatedly learned fastest. He uses the experience as an analogy for the persistence and experimentation required in company building.

He regrets spending on a separate special-projects team. In his view, the core organization should be able to innovate, and people who discover a promising idea should have the chance to own and scale it rather than hand it off.

Spain is a difficult market for Choco because restaurant supply is fragmented, he says. He estimates an average restaurant orders from about 4.5 suppliers in the US or Germany, nine in France and 18 in Spain; the numbers are his operational estimates.

With hindsight, he wishes the company had invested more aggressively in its first two years once its product and go-to-market economics were working. He distinguishes scaling a proven model from simply increasing burn because competitors raised money.

He credits a well-funded early competitor with pushing Choco to analyze strengths and weaknesses and improve. Ultimately he prefers intrinsic motivation around the company’s mission to defining success solely by beating a rival.

In the quick-fire round, Khachab predicts that AI-first companies will win and that using AI internally will become ordinary business practice. This is his forecast; he defines AI-first in part by the share of revenue generated from AI products.

He declines to give revenue but says Choco’s GMV reached roughly $1.5 billion after around five years, then added another $1 billion in less than a year. GMV measures goods traded through the platform and is not company revenue; the figures are his account.

He would like to spend a day at Pixar to learn about creativity and talent, and at Deutsche Bahn to understand what he sees as operational problems. His criticisms of the railway and its bonus arrangements are his claims, not findings established in the interview.

He notes that the same fast action may be praised as decisive or criticized as impulsive depending on the outcome. He also predicts that automation will make some great companies smaller by headcount and that a very small team could build a billion-dollar company.

The question he wants investors to ask is why a founder is personally willing to commit for many years. He values competition and excellence but treats valuation as a result of the work rather than the work itself.

The conversation turns to public infrastructure and living standards in Silicon Valley. Khachab questions why prosperous technology hubs have not solved problems such as transport and homelessness; this is his observation, not a measured assessment of every outcome.

He argues that businesses should align economic success with positive social or environmental effects, and says Choco’s mission includes reducing food waste. The extent of the company’s impact is not independently measured in the interview.

Khachab says he sometimes eats meat and argues attention should go to interventions with the greatest impact. He claims reducing food waste would have much greater emissions impact than eliminating cars; the comparison is his and the interview does not establish its methodology.