Microsoft 365 Copilot: What Is Included at Launch and What It Costs

Microsoft 365 Copilot costs 30 US dollars per user per month and has been available to business customers since 1 November 2023, initially for customers with Microsoft 365 E3 or E5 and from 300 licences. The assistant works in the office applications from Word to Teams and draws on the data that each user is allowed to see in the Microsoft 365 tenant. Whether the licence pays off is decided less by the technology than by the roles that receive it, on how tidy the permissions are and on whether the time gained is actually used differently in daily work.

This article describes what Microsoft offers at launch, works through the cost for a company and sets out criteria for deciding who should receive a licence first. It touches on data protection only in passing: this article provides an overview and does not constitute legal advice.

What is included in Microsoft 365 Copilot at launch?

Microsoft 365 Copilot brings an AI assistant into the applications where office work takes place. Microsoft announced availability on 21 September 2023 and set general availability for business customers on 1 November 2023. The assistant appears in the Microsoft 365 applications, from Word and Excel to Teams and OneNote. In Word it drafts text and summarises documents, in Outlook it condenses long email threads, in Teams it summarises meetings, in PowerPoint it produces a first slide draft from a document, and in Excel it helps with the analysis of spreadsheets.

Added to this is Microsoft 365 Chat, a chat window that brings together information from the user's documents, messages and meetings. Microsoft describes the processing as taking place within the organisation's own Microsoft 365 tenant, and according to the vendor the assistant respects existing permissions. That is an advantage over freely available chatbots, which employees often use outside any control. It is also the reason why the permission structure must be reviewed before rollout.

Bing Chat Enterprise is a separate product: a chat assistant with data protection commitments for business customers. In July 2023 Microsoft announced that it would include it at no extra charge in Microsoft 365 E3, E5, Business Standard and Business Premium and later offer it as a standalone product for 5 US dollars per user per month. It answers questions from the web and has no access to company data. For organisations that first want to gain experience with generative AI, it is the cheaper entry point. How decision-makers can assess vendor announcements is discussed in the article on the AI hype cycle.

What does Microsoft 365 Copilot cost?

Microsoft 365 Copilot costs 30 US dollars per user per month, on top of the Microsoft 365 licence the user already holds. On an annual basis that is 360 US dollars per user, roughly 335 euros at an exchange rate of about 1.07 US dollars to the euro. Two restrictions apply at launch: the licence is open to customers with Microsoft 365 E3 or E5, and the minimum purchase is 300 licences. That comes to around 108,000 US dollars or roughly 100,000 euros a year. Microsoft has announced Business Standard and Business Premium for the licence, but they are not included at launch. Euro prices are settled with the Microsoft partner, because the terms may differ from the US dollar price list.

The licence is only one part of the cost. Added to it are the effort of cleaning up permissions and of training users, as well as the business units' time for the pilot and its evaluation. As a rule of thumb, a company should budget for enabling its users alongside every licence, because an assistant without a habit of use rarely saves time.

What do studies show about generative AI at work?

Studies of generative AI at work show noticeable time gains in writing and knowledge work, above all for people early in their careers. At launch there are no independent studies of Copilot itself, but there are studies of comparable assistants built on large language models. An experiment with professionals, published in the journal Science in 2023, had participants complete occupational writing tasks with and without ChatGPT: the time required fell by around 40 per cent, and the quality of the results, as rated by third parties, rose by around 18 per cent. A field study of more than 5,000 customer service agents, published in 2023 as a working paper of the National Bureau of Economic Research, found on average 14 per cent more issues resolved per hour, and 34 per cent more among novices.

Two caveats belong to the reading of these figures. The tasks were clearly defined, and the results apply to writing and customer dialogue, not to every kind of office work. The service study also shows that the effect was markedly smaller for experienced agents. Decision-makers should conclude that the benefit varies with the task and with the experience of the user, and has to be measured in their own pilot. A sober assessment of the first year of generative AI in companies is given in the article on ChatGPT in the company.

Who should receive a licence first?

Licences go first to roles with a lot of text and many meetings, where results are reviewed before they are passed on anyway. What counts is the task someone performs every day, not their position. The following table shows a selection of typical roles with a hypothesis about the benefit that the pilot must confirm or refute.

How do you roll out Microsoft 365 Copilot in four steps?

You roll out Copilot by equipping a pilot group with defined tasks and cleaning up permissions beforehand. You then measure the benefit and expand only after that. The pilot ends in a decision and does not run on as a permanent state. Because of the minimum purchase of 300 licences, one precondition applies: whoever buys Copilot at launch books a pilot the size of the minimum. Organisations with fewer than 300 users, or without E3 or E5, first gain experience with Bing Chat Enterprise, which is included in their licences, and wait for the announced opening of the licence.

1. Set the goal and the baseline: the executive team decides which metric should change, for example the time spent on meeting follow-up or the lead time of quotations. The baseline is collected over two to four weeks before the start.
2. Order the permissions: the assistant shows what users are allowed to see. Wrongly shared folders and documents that have accumulated in many companies over the years become visible at once. IT therefore checks before the start which sensitive content is shared widely.
3. Equip the pilot group: the 300 licences go to users from three to four roles with the highest expected benefit, each with an introduction to working with the assistant and a point of contact. Usage is logged, and users report weekly on tasks where the assistant helped or got in the way.
4. Decide: after eight to twelve weeks the executive team compares the metric with the baseline, sets it against the cost and decides for which roles the licence is extended. How pressure to adopt AI becomes such a sequence of decisions is described in the article on AI activism.

What risks should decision-makers know?

The main risks are faulty content, inflated expectations and disordered data. Language models phrase things convincingly and are sometimes wrong, so every draft that leaves the company belongs in the hands of a person who answers for its content. Inflated expectations arise when demonstrations show the best cases while everyday work involves incomplete documents and special cases. Disordered data means that the assistant finds information that no employee would deliberately have looked for. There is also an organisational risk: if the licence is allocated to everyone at once, the attention of leadership is spread thin and nobody builds routines.

Microsoft 365 Copilot with Niehues Consulting

Niehues Consulting is a Düsseldorf consultancy for AI management and organisation design from the delivery perspective. Founder Marin Niehues brings AI into operation, in projects ranging from VC-funded scale-ups to DAX 40 groups and in more than 20 countries on six continents. Within AI management, the work runs from business analysis and requirements through the business case to the selection and sequencing of applications and portfolio steering.

The approach follows the Niehues Delivery Model with its three phases. Three of its review questions matter most when introducing an assistant such as Copilot: Which business result should change measurably? Is the business case confirmed with real values? Are the users enabled and the processes adapted? The scope and terms of an engagement are discussed in the free initial consultation of 30 minutes, which you can arrange via contact.