AI service providers mine a rich seam of demand for efficiencies

The business market for artificial intelligence (AI) software is booming as companies seek new ways to cut costs and auto­mate laborious tasks. And that trend is already reflected in this year’s FT-Statista ranking of the Americas’ 500 fastest-growing companies: it includes at least two dozen that claim to use AI or machine learning in their products.

The market for AI services is estimated to be worth $50bn in the US alone this year, but it will still be between three to five years before a majority of companies are using the technology, predicts Nick Patience, managing analyst at 451 Research, part of S&P Global Market Intelligence.

Here, the FT examines three of the AI companies in this year’s ranking.

Automated ads

Waymark makes television adverts for clients based on what is freely available on the internet. The Detroit-based company, which launched in 2017, reported revenue of $4.1mn in 2022, up from $2.4mn in 2019, placing it 442nd in the ranking.

The company says it can produce 30-second video adverts, with an AI voiceover, within minutes. Customers provide brief instructions and then Waymark’s software — which uses 13 AI models, including OpenAI’s ChatGPT — imports data from the client’s website, such as marketing information and logos, as well as from social media and local business databases.

Waymark’s software produces two or three draft versions of the ad. The customer picks one and edits it before publication. Prices start at $150 for two video ads a month.

“Smaller ad agencies love us [because they use our software] to create more ads for their customers [who were] previously unable to afford ads,” says Waymark chief executive, Alex Persky-Stern.

US radio group NRG Media has used Waymark’s AI software to create a video advertising campaign. The ads can be finished in about 30 minutes, compared with hours previously, says Erica Dreyer, director of integrated media at NRG Media. “We have found AI gets us about 80 per cent of the way there and then we can do some tweaks [to the ad],” she says.

Previously, a need to shoot videos on location, and a lack of onsite videographers, limited the number of ads NRG Media could produce, she says.

Waymark developed its video-generation software before generative AI “was a big thing”, says Persky-Stern. The company has since added generative AI features to its product. “Our founder is a big AI person and as soon as [OpenAI’s] GPT-3 was released in beta form [in 2020], we got access and started developing against it for our video generation software,” he explains.

Data labelling

Like any business technology, an AI system is only as good as the quality of data fed into it. This is where Cogito Tech comes in. It labels the vast amount of data − images, videos, audios and text — that AI systems need to consume, and learn from, in order to ensure their output is high quality. It also sources reliable data, such as media articles, for its corporate clients’ AI systems.

If, for example, a manufacturing company wanted its assembly-line AI systems to be trained to detect cracks or dents in its products, it could provide examples to Cogito’s data analysts. They would then verify and categorise the data, ensuring it is of sufficient quality to be used to train the manufacturing company’s AI system.

“The biggest thing with AI is: junk in, junk out,” says Rohan Agrawal, Cogito chief executive. “If that data is not of good quality, then you basically end up training a very, very biased or bad model.”

Founded in 2011, Cogito is benefiting from a surge in demand for AI and the “large language models” that underpin systems such as ChatGPT and Google’s Gemini. Cogito has about 2,000 employees in New York and Uttar Pradesh, northern India. Around 70 per cent of its business is in the US, with the remainder in Europe. In 2022, Cogito’s revenue was $8.9m − up from $2mn in 2019, ranking it 137th in the FT list.

Its customers are typically companies with between $500mn and $2bn in revenue. Many will have large IT departments, but are unable to prepare their own AI data because the labelling work is “tedious” and “labour intensive”, explains Agrawal. Companies would rather focus on building the AI models, he says. “During the gold rush, you sell shovels; we’re doing the same thing for AI,” Agrawal argues. “We are basically equipping [companies] with data, which is the new oil . . . to build highly effective AI models. And, as far as we are concerned, we [are] just scratching the surface.”

Legal services

AI that can automate data collection and analysis is also much in demand — and legal services providers have been early adopters. Over the past decade, AI has automated many laborious legal tasks, such as reviewing long contract documents to find clauses that contravene legal policy, and to ensure compliance with financial and data regulations.

LinkSquares specialises in AI software for companies’ legal departments. Ranked 43rd in the FT-Statista table, it reported revenue of $24mn in 2022, up from $2mn in 2019.

Founded in 2015 by two non-lawyers, the Boston-based company now helps organisations including Time magazine and restaurant chain TGI Fridays to save time, keep track of documents, and manage projects.

“[Our AI software] might analyse a large contract and suggest where certain clauses could be changed to make them more favourable to a client,” says Vishal Sunak, LinkSquares’ founder and chief executive. Such automation can halve the time it takes companies to draft and sign contracts, the company says.

Much of its growth is from company legal teams using AI for the first time, says Sunak. “There’s so much . . . greenfield market opportunity. Businesses now understand that they need technologies like [our AI] to . . . work more efficiently and work more collaboratively.”