AI in the First Half of 2024: Which Announcements Have Reached Everyday Work

What has reached everyday work in the first half of 2024 are mainly assistants for writing and programming that are built into existing tools. Video generation and speaking assistants remained announcements or previews. This article places the major news of the first six months on four maturity levels and gives you a checklist for judging any new announcement for your company.

The pace of announcements was high this half-year. In February OpenAI showed Sora, a model that creates short videos from text. In March Anthropic presented the Claude 3 model family, and a start-up demonstrated Devin, an AI software engineer. May brought GPT-4o from OpenAI, Google's developer conference with Gemini 1.5 and AI overviews in search, and Microsoft's Copilot+ PC. In June Apple announced Apple Intelligence at its developer conference, and Anthropic released Claude 3.5 Sonnet. For an executive team every one of these items raises the same question: must we act now?

How do you tell whether an AI announcement has reached everyday work?

An AI announcement has reached everyday work when employees use it without special access in their familiar tool and the effect on a metric is documented. Four maturity levels make this testable.

The levels rest on vendor statements and on publications up to the end of the half-year. The placement of an item can change, which is why the check belongs in the quarterly round of AI steering.

Which announcements of the first half-year are available, and which are not?

Available are the language models and the assistants integrated into office tools; not yet available are video generation, AI software engineers and the speaking assistant. The placement in detail:

Sora, announced in February, was not publicly accessible at the end of the half-year and stands at level 1. The article on Sora and generative video describes what follows for companies. Devin, demonstrated in March, stood at level 1 to 2 with restricted access. Apple Intelligence, announced at the developer conference in June, had not yet been delivered and stands at level 1. The consequences for corporate IT are described in the article on Apple Intelligence.

GPT-4o, presented in May, stands at level 3 for text and images. The announced voice function with real-time dialogue, by contrast, was not generally released at the end of the half-year and stands at level 2. How a speaking assistant can be placed in customer contact is described in the article on GPT-4o in customer service. The assistant in Microsoft's office tools has been generally available to companies since November 2023 and stands at level 3. The AI overviews in Google search were rolled out in the United States in May and were then publicly criticised for faulty answers in individual cases, and the provider made corrections. For companies the case shows that even available functions stay at level 3 as long as their reliability is not documented.

Two groups stand at level 4. The first is assistants in writing work and customer service. An experiment with 453 professionals found a completion time about 40 per cent shorter with 18 per cent higher rated quality, and a field study of more than 5,000 support agents found 14 per cent more issues resolved per hour. The second group is programming assistants: in a controlled trial, developers with an assistant completed a task on average 55.8 per cent faster than the comparison group.

How many companies actually use AI?

According to the statistical office of the European Union, 8 per cent of enterprises with at least ten employees used AI technologies in 2023. The figure was published at the end of May 2024. It shows that attention to AI goes far beyond actual use, and it explains why the impression of falling behind is often larger than the gap itself. For products that call themselves AI, the first test question is therefore which task they demonstrably improve. The article on recognising AI washing deepens it.

How do you assess a new announcement in five questions?

You assess a new announcement with five questions, each of which must be answered in one sentence before budget flows.

1. At which maturity level does the announcement stand, and who has documented that placement?
2. Which task in your company would it change, and which metric measures that?
3. Does it fit the tools your employees use every day today?
4. What does it cost over twelve months, including integration and operation?
5. Who in the business unit is willing to own a test with a clear stop criterion?

If the answer to question 1 is “level 1”, the item belongs on the watch list, and the next review date is the end of the quarter. If it is “level 3”, a pilot with a clear hypothesis follows. Only at level 4, confirmed by your own measurement, does the rollout begin.

What does this mean for the second half of the year?

A rule for the second half of the year can be drawn from the first: invest in spreading what stands at levels 3 and 4, and keep a small learning budget for levels 1 and 2. The applications with documented effect lie in writing work, in programming and in customer service. Whoever deploys them in core processes gains more than by waiting for the next model generation. A learning budget for new functions keeps the organisation agile without taking on operational responsibility for immature solutions.

What matters is that every change of level becomes visible in the organisation. A simple watch list with announcement and level as well as owner and next review date is enough. It prevents functions from triggering new discussions on a continuous basis, and it gives the executive team an answer when the next report of a supposed breakthrough arrives.

AI decisions with Niehues Consulting

Niehues Consulting is a consultancy for AI management and organisational design from a delivery perspective, based in Düsseldorf. Founder Marin Niehues brings AI into operation, in projects from VC-funded scale-ups to DAX 40 groups and in more than 20 countries on six continents. In the AI management service area the work covers business analysis, the business case, and the selection and sequencing of applications, as well as portfolio steering.

The approach follows the Niehues Delivery Model with three phases. Three of its test questions matter most for assessing announcements: which business outcome should change measurably? Does the prototype prove the benefit in the real process with real data? Which metric shows after 90 days that the solution is landing? Scope and terms of an engagement are discussed in the free initial consultation, which you can arrange in 30 minutes through the contact form.