SAN FRANCISCO/STOCKHOLM, Dec 16 – Final spring, CellarTracker, a wine-collection app, constructed an AI-powered sommelier to make unvarnished wine suggestions primarily based on an individual’s palate. The issue was the chatbot was too good.
“It is simply very well mannered, as an alternative of simply saying, ‘It is actually unlikely you may just like the wine,’” CellarTracker CEO Eric LeVine mentioned. It took six weeks of trial and error to coax the chatbot into providing an trustworthy appraisal earlier than the characteristic was launched.
Enroll right here.
Since ChatGPT exploded three years in the past, firms large and small have leapt on the likelihood to undertake generative synthetic intelligence and stuff it into as many merchandise as potential. However thus far, the overwhelming majority of companies are struggling to appreciate a significant return on their AI investments, in line with firm executives, advisors and the outcomes of seven latest government and employee surveys.
One survey of 1,576 executives performed throughout the second quarter by analysis and advisory agency Forrester Analysis confirmed simply 15% of respondents noticed revenue margins enhance as a result of AI over the past yr. Consulting agency BCG discovered that solely 5% of 1,250 executives surveyed between Might and mid-July noticed widespread worth from AI.
Executives say they nonetheless imagine generative AI will ultimately rework their companies, however they’re reconsidering how rapidly that may occur inside their organizations. Forrester predicts that in 2026 firms will delay about 25% of their deliberate AI spending by a yr.
“The tech firms who’ve constructed this expertise have spun this story that that is all going to vary rapidly,” Forrester analyst Brian Hopkins mentioned. “However we people don’t change that quick.”
AI firms together with OpenAI, Anthropic and Google are all doubling down on courting enterprise clients within the subsequent yr. Throughout a latest lunch with media editors in New York, OpenAI CEO Sam Altman mentioned growing AI techniques for firms may very well be a $100 billion market.
All that is taking place in opposition to the backdrop of unprecedented tech funding in every thing from chips, to knowledge facilities, to power sources.
Whether or not these investments will be justified can be decided by firms’ potential to determine how you can use AI to spice up income, fatten margins or velocity innovation. Failing that, the infrastructure build-out may set off the sort of crash harking back to the dot-com bust within the early 2000s, some specialists say.
THE ‘EASY’ BUTTON
Quickly after ChatGPT’s launch, firms worldwide created activity forces devoted to discovering methods to embrace generative AI, a kind of AI that may create authentic content material like essays, software program code and pictures by means of textual content prompts.
One well-known subject with AI fashions is their tendency to please the person. This bias – what’s referred to as “sycophancy” – encourages customers to speak extra, however can impair the mannequin’s potential to provide higher recommendation.
CellarTracker bumped into this downside with its wine-recommendation characteristic, constructed on high of OpenAI’s expertise, CEO LeVine mentioned. The chatbot carried out effectively sufficient when requested for basic suggestions. However when requested about particular vintages, the chatbot remained optimistic – even when all indicators confirmed an individual was extremely unlikely to get pleasure from them.
“We needed to bend over backwards to get the fashions (any mannequin) to be vital and counsel there are wines I won’t like,” LeVine mentioned.
A part of the answer was designing prompts that gave the mannequin permission to say no.
Corporations have additionally struggled with AI’s lack of consistency.
Jeremy Nielsen, basic supervisor at North American railroad service supplier Cando Rail and Terminals, mentioned the corporate lately examined an AI chatbot for workers to review inside security experiences and coaching supplies.
However Cando ran right into a stunning stumbling block: the fashions couldn’t persistently and accurately summarize the Canadian Rail Working Guidelines, a roughly 100-page doc that lays out the security requirements for the trade.
Generally the fashions forgot or misinterpreted the principles; different occasions they invented them from complete fabric. AI researchers say fashions usually battle to recall what seems in the course of an extended doc.
Cando has dropped the venture for now, however is testing different concepts. To this point the corporate has spent $300,000 on growing AI merchandise.
“All of us thought it’d be the simple button,” Nielsen mentioned. “And that’s simply not what occurred.”
HUMANS MAKE A COMEBACK
Human-staffed name facilities and customer support had been speculated to be closely disrupted by AI, however firms rapidly discovered there are limits to the quantity of human interplay that may be delegated to chatbots.
In 2025, nonetheless, CEO Sebastian Siemiathowski was pressured to dial that again and acknowledge that some clients most well-liked to speak with people.
Siemiathowski mentioned AI is dependable on easy duties and may now do the work of about 850 brokers, however extra advanced points rapidly get referred to human brokers.
For 2026, Klarna is targeted on constructing its second-generation AI chatbot, which it hopes to ship quickly, however human beings will stay an enormous a part of the combination.
“If you wish to keep customer-obsessed, you may’t rely [entirely] on AI,” he mentioned.
Equally, U.S. telecommunications large Verizon is leaning again into human customer support brokers in 2026 after makes an attempt to delegate calls to AI.
“I believe 40% of customers like the concept of nonetheless speaking to a human, they usually’re annoyed that they cannot get to a human agent,” mentioned Ivan Berg, who leads Verizon’s AI-driven efforts to boost service operations for enterprise clients, in a Reuters interview this fall.
The corporate, which has about 2,000 frontline customer support brokers, nonetheless makes use of AI to display calls, get info on clients, and direct them to both self-service techniques or to human brokers.
Utilizing AI to deal with routine questions frees up brokers to deal with advanced points and take a look at new issues, equivalent to making outbound calls and doing gross sales.
“Empathy might be the important thing factor that is holding us from having AI brokers discuss to clients holistically proper now,” Berg mentioned.
Shashi Upadhyay, president of product, engineering and AI at customer-service platform Zendesk, says AI excels in three areas: writing, coding and chatting. Zendesk’s shoppers depend on generative AI to deal with between 50% and 80% of their customer-support requests. However, he mentioned, the concept generative AI can do every thing is “oversold.”
THE ‘JAGGED FRONTIER’
Massive language fashions are quickly conquering advanced duties in math and coding, however can nonetheless fail at comparatively trivial duties. Researchers name this contradiction in capabilities the “jagged frontier” of AI.
“It could be a Ferrari in math however a donkey at placing issues in your calendar,” mentioned Anastasios Angelopoulos, the CEO and cofounder of LMArena, a well-liked benchmarking software.
Seemingly small points can unexpectedly journey up AI techniques.
Many monetary companies depend on knowledge compiled from a broad vary of sources, all of which will be formatted very in a different way. These variations may immediate an AI software to “learn patterns that don’t exist,” mentioned Clark Shafer, director at advisory agency Alpha Monetary Markets Consulting.
Many firms at the moment are wanting into the possibly costly, prolonged and sophisticated technique of reformatting their knowledge to benefit from AI, Shafer mentioned.
Dutch expertise funding group Prosus says one in every of its in-house AI brokers is supposed to reply questions on its portfolio, just like what the group’s knowledge analysts on workers already do.
Theoretically, an worker may ask how usually a Prosus-backed food-delivery agency was late to ship sushi orders in Berlin final week.
However for now, the software doesn’t all the time perceive what neighborhoods are a part of Berlin or what “final week” means, mentioned Euro Beinat, head of AI for Prosus.
“Individuals thought AI was magic. It isn’t magic,” Beinat mentioned. “There’s a variety of information that must be encoded in these instruments to work effectively.”
MORE HANDHOLDING
OpenAI is engaged on a brand new product for companies and lately created inside groups, such because the Ahead Deployed Engineering staff, to work instantly with shoppers to assist them use OpenAI’s expertise to deal with particular issues, a spokesperson mentioned.
“The place we do see failure is those who soar in too large, they discover that billion-dollar downside—that is going to take a number of years,” mentioned Ashley Kramer, OpenAI’s head of income, throughout an onstage interview at Reuters Momentum AI convention in November.
Particularly, OpenAI is working with firms to seek out areas the place AI can have a “excessive affect however possibly low carry at first,” mentioned Kramer.
Rival AI lab Anthropic, which attracts 80% of its income from enterprise clients, is hiring “utilized AI” specialists who will embed with firms.
For AI firms to succeed, they should view themselves as “companions and educators, fairly than simply deployers of expertise,” mentioned Mike Krieger, Anthropic’s head of product, in an interview earlier this yr.
An growing variety of startups, many based by former OpenAI staff, are growing AI instruments for particular sectors equivalent to monetary providers or authorized. These founders say firms will profit from specialised fashions greater than general-purpose or client instruments like ChatGPT.
It’s a playbook that Author, a San Francisco–primarily based AI software startup, has been adopting. The corporate, which is now constructing AI brokers for finance and advertising groups at giant companies equivalent to Vanguard and Prudential, places its engineers on calls instantly with shoppers to grasp their workflows and co-build the brokers.
“Corporations want extra handholding in really making AI instruments helpful for them,” mentioned Might Habib, CEO of Author.
Reporting by Deepa Seetharaman and Krystal Hu in San Francisco and Supantha Mukherjee in Stockholm. Enhancing by Kenneth Li and Michael Learmonth.
Our Requirements: The Thomson Reuters Belief Ideas.
Leave a Reply