For most of its 48 years, Microsoft has been known as the tech industry’s most effective fast-follower — a company that perfected inventions made elsewhere, rather than coming up with its own.
But this week has brought something different, according to Barry Briggs, a former chief technology officer at the US company: Microsoft has become the first in the industry to make the technology behind ChatGPT available as a standard feature in a widely used software product, potentially transforming the working lives of millions.
After nearly a year of soaring hopes for generative AI, the software industry will be watching the outcome for signs of whether the technology is ready to move from hype to business reality. Microsoft’s move will also be the first test of companies’ willingness to pay a high price for a technology that draws heavily on expensive computing resources such as the advanced chips needed to train AI models, and which make it costly to deliver.
On Wednesday, Microsoft officially declared “general availability” of a generative AI assistant, dubbed Copilot, in enterprise versions of its widely used Microsoft 365 suite of productivity apps, which includes Word, PowerPoint and Excel.
The move potentially puts new AI tools at the fingertips of an estimated 150mn workers, according to analysts, and helping them automatically generate documents and emails or create spreadsheets more easily. The software is designed to make it simple to draw on all the data a company holds in its Microsoft applications. Eventually, connections to other data stores are meant to make Copilot a “smart” front-end for working with all of a company’s most valuable data.
To judge from the price tag, Microsoft expects the technology to mark a profound advance in the everyday software applications used by many white-collar workers. The Copilot feature has been priced at $30 a month per user and is available as an upgrade to customers who already pay for the E3 or E5 editions of the company’s productivity apps. These cost $36 and $57 a month, respectively — meaning that the upgrade will add as much as 83 per cent to the monthly software bill.
“At that price point customers are going to be reluctant to adopt it enterprise-wide straight away,” said Derrick Wood, an analyst at TD Cowen. Instead, he added, they are likely to test it carefully before rolling it out, searching for the particular types of work that benefit the most from the technology.
Despite this, the sheer number of workers who use its software and its early lead in generative AI has put Microsoft in pole position as the fast-evolving technology makes it into everyday use.
At a recent Gartner event attended by 9,000 corporate chief information officers and others, Microsoft’s AI dominated much of the conversation, said Jason Wong, an analyst at the IT research firm. One early sign of that came last week with an unexpected pick-up in growth at Microsoft’s Azure cloud platform, which the company attributed to AI.
“Microsoft has done well to generate mind share,” said Wong. “But it is not market share yet.” Other software companies “will be watching very closely” to see if it can turn the promise into real revenue gains, he added.
Unlike a normal product launch, there is much about the arrival of generative AI in mass-market software that looks experimental. It is not yet clear what types of work will benefit most from the technology or how easily workers will take to the new tools, analysts say.
There are also “risks of errors and misinterpretations and misuse”, said JP Gownder, an analyst at Forrester Research. A lazy worker, for instance, might send an AI-generated email without checking it for accuracy, or rely on an automated recap of a meeting without checking with one of the humans who was there to pick up nuances the AI failed to grasp. If AI makes it easier to produce more emails, it may add to the overload felt by many workers who are already drowning in electronic communications.
Microsoft executives admit there are many things they don’t know about how workers will use the new technology or what unexpected side-effects it may have.
“Certainly we’re in the early innings and we’re just learning,” said Jared Spataro, the Microsoft corporate vice-president in charge of bringing new features to its business applications. “I’m sure there are things we’re not thinking about right now.” But he predicted that positive surprises would outweigh the negative ones as the versatility of generative AI brings new and unexpected uses.
Spataro added that Microsoft has already collected enough data from its trials to show the new Copilot delivers big productivity gains. It plans to disclose the findings at its annual Ignite conference on November 14.
Besides measuring the time workers save by doing tasks such as automatically generating text and analysing numbers, Microsoft says its tests have included running control groups to validate the findings and developing ways to measure the quality of the generative AI output, rather than just its quantity.
Until customers see the results and get a chance to validate them with their own tests, however, such claims have to be taken on trust. “Right now, there aren’t a lot of facts,” said Wong. “Companies are either investing on faith, or fear: fear they’ll fall behind.”
The $30 a month figure was set after researching what customers would be prepared to pay for the sort of assistance the AI can deliver, said Spataro. Microsoft also looked at the “cost-per-head math” — how much $30 a month represented in terms of the overall cost of hiring a knowledge worker, and what kind of productivity gains would be needed to justify the extra software cost.
Gownder at Forrester calculated that the extra spending would be justified even with “pretty modest assumptions”, for instance assuming the software saved a high-priced worker as little as two hours a month.
Despite making big claims for the AI, Microsoft executives have sought to lower expectations about its revenue impact in the short term, and Wall Street forecasts are tempered. Even a successful rollout of the technology might only add 1 per cent to Microsoft’s revenue growth in its 2025 fiscal year, which starts next July, calculates TD Cowen analyst Wood. That is the equivalent to around $2-2.5bn, Wood adds.
Competition with other software companies could also limit the potential. The APIs offered by OpenAI and other makers of large language models, which make it easy for programs to tap into their services, have opened the way for generative AI to become a standard feature of many software applications.
Microsoft will still have a number of advantages over other companies, says Wood, such as being able to tap into data held in Microsoft applications and its close relationship with OpenAI. But it is not clear how much of a premium it will be able to command if automatically generating emails or summarising documents become routine activities available from many different applications. “We’ll see if Microsoft can hold these price points,” he added.
Read the full article here
