# Apple Intelligence: What AI on the Smartphone Means for Enterprise IT

Author: Marin Niehues, Niehues Consulting · Published: 2024-06-13 · AI Strategy & Business Case
Source: https://www.niehuesconsulting.de/en/blog/apple-intelligence-smartphone-enterprise-it/

Apple Intelligence brings generative AI directly to iPhone, iPad as well as Mac, and so into the hands of employees who need no IT project to use it. For enterprise IT this means steering usage before it begins: with a classification of data, a check of the device fleet and a limited pilot. The announcement of 10 June 2024 gives enough pointers to start planning now.

This article summarises what Apple announced at its WWDC developer conference, shows which questions enterprise IT should ask and offers a checklist for preparation. This article provides an overview and does not constitute legal advice.

**When AI becomes a feature of the operating system, IT sets the rules under which it is used and no longer selects a tool. Whoever settles the rules before availability gains time.**

## What did Apple announce with Apple Intelligence?

On 10 June 2024 at WWDC, Apple announced an AI system for iOS 18, iPadOS 18 as well as macOS Sequoia that runs language models on the device and, where needed, on Apple's own servers. The features cover writing, images and Siri. The writing tools rewrite, proofread and summarise text in Mail, Notes and other applications. Notifications are sorted by urgency and summarised. According to the announcement, Siri understands personal context, such as messages and appointments, and can carry out actions inside applications.

Apple sets two conditions. The features run only on devices with enough computing power: on the iPhone 15 Pro and iPhone 15 Pro Max as well as on iPad and Mac with the M1 chip or later. In addition, according to Apple they launch in autumn 2024 as a beta, initially in US English, with further languages and features to follow over the course of the next year. Older devices in the fleet remain excluded.

## How does Private Cloud Compute work?

Private Cloud Compute is Apple's method of running more demanding requests on servers with its own chips, without the data being stored or seen by Apple. According to Apple's description, the device processes as much as possible itself and sends the servers only the data a given request needs. Apple states that independent experts can inspect the server code. For decision-makers this is a vendor's promise whose effectiveness is shown only by independent review and by documentation.

A third processing level involves ChatGPT. With the user's consent, Siri can pass requests to OpenAI's ChatGPT without an account being required, and the writing tools can use it as well. For IT this means three data paths on one device: local, via Apple's servers and via an external provider. Each path needs its own assessment.

## Which questions must enterprise IT settle now?

Enterprise IT must settle five questions: which devices are affected, which data may be processed, whether the features can be controlled, who should use them and how the benefit is measured. The first three are IT tasks, and the last two belong to the business units.

**Five questions on Apple Intelligence in the enterprise**
Criterion | Question | What to do | Responsible
Device fleet | How many devices meet the hardware requirement? | Analyse the inventory by model, review replacement cycles | IT operations
Data classes | Which data may an employee enter into the writing tools and Siri? | Define data classes, name what is prohibited | Information security
Controllability | Can features and the path to ChatGPT be managed centrally? | Check device management documentation once available | IT operations
Benefit | Which tasks become faster or better? | Set up a pilot group with metrics | Business unit
Rules | What does the usage policy say? | Adopt a policy before release | Executive management

The third question cannot be answered conclusively at the time of the announcement. Apple said nothing about management by enterprises in the announcement, and documentation on it is still to come. Enterprise IT should therefore record in advance what it expects from management: features that can be switched off individually, a separately blockable path to ChatGPT as well as logging of usage. Data protection and information security review each of these requirements case by case, and legal assessment lies with legal specialists.

## What changes for employees?

For employees, AI features appear in the everyday tools they already use, in Mail, Notes, Messages as well as Phone. The hurdle of opening a separate AI tool disappears. That lowers the barrier to entry and raises the risk that confidential content is processed unintentionally. A field sales employee who polishes a customer email with the writing tool makes a data decision without noticing.

A simple measure follows from this: a one-page policy that says in examples what is permitted. Two examples per data class work better than a paragraph of principles. How teams handle parallel AI tools in practice and when this turns into sprawl is described in the article on [AI subscription sprawl](/en/blog/ai-subscription-sprawl/).

## Does Apple Intelligence change the competitive position of the vendors?

Apple Intelligence mainly changes the reach of AI features, because they arrive on devices that are widespread in companies. The combination of local processing, own cloud and an external partner resembles the approach of other vendors. In office software Microsoft has taken a comparable route with [Microsoft 365 Copilot](/en/blog/microsoft-365-copilot/), where AI is integrated into the applications. For companies this creates a platform decision: which AI features come from the operating system, which from the office suite and which from applications the company builds itself?

Whoever does not take this decision has it taken by the vendors. A simple grid is sensible that assigns each feature to one of the three sources and names owners per source. Preparing for new announcements always follows the same logic, described in the article on the [AI hype cycle](/en/blog/ai-hype-cycle-progress-versus-announcement/): announcement, availability and benefit are three different events.

## How do you plan a pilot for Apple Intelligence?

You plan a pilot in four steps: choose the group, limit the data, set the metrics and decide after eight weeks. The pilot group comprises ten to fifty people with suitable devices and typical writing and communication tasks, for example in sales, customer service or project coordination. The data rules are binding for the group. What is measured is time per task, quality of results as well as user satisfaction, each against a baseline taken before the pilot.

After eight weeks executive management decides on the basis of the measurement whether the features are released to further groups, under which rules and with what training need. Without a baseline this decision has no evidence behind it, so the measurement comes before the release of the features. How pressure becomes a robust plan is described in the article on [AI activism](/en/blog/ai-activism-board-pressure-into-plan/).

**Checklist: Preparing for Apple Intelligence**
- Device inventory analysed by model and operating system?
- Data classes defined and backed with examples?
- Requirements for device management formulated?
- One-page usage policy written?
- Pilot group named with baselines and metrics?
- Owner for data protection and information security involved?
- Decision date for release fixed?

## AI on the smartphone with Niehues Consulting

[Niehues Consulting](/en/) is a consultancy for AI management and organisation design from a delivery perspective, based in Düsseldorf. Founder Marin Niehues brings AI into operations, in projects ranging from VC-funded scale-ups to DAX 40 corporations and in over 20 countries on six continents. In the service area of 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. Two of its review questions matter for features such as Apple Intelligence: which business outcome should change measurably? And: 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 via the [contact page](/en/#kontakt).

**Conclusion**
Apple Intelligence moves AI into the operating system and thus into employees' everyday tools. Enterprise IT now defines data classes, device requirements as well as a policy, and after availability runs a pilot with metrics. Whoever is prepared decides on release on the basis of data.

**Sources**
- Apple: Apple Intelligence and Private Cloud Compute, press release for WWDC, 10 June 2024

## Frequently asked questions

### What is Apple Intelligence?

Apple Intelligence is the AI system Apple announced on 10 June 2024 for iOS 18, iPadOS 18 as well as macOS Sequoia. It comprises writing tools, image features and an extended Siri that uses language models on the device and in Apple's own cloud. Rollout starts in autumn 2024 as a beta.

### Which devices support Apple Intelligence?

According to Apple, the iPhone 15 Pro and iPhone 15 Pro Max as well as iPads and Macs with the M1 chip or later support it. Older devices do not receive the features. For companies this means analysing the device inventory by model.

### Is Apple Intelligence safe for business use?

Apple describes a method in which as much as possible is processed on the device and server requests run without storage. Whether that suffices for a company is answered by an information security review against the documentation. Passing requests to ChatGPT happens with the user's consent and needs its own assessment.

### Can IT manage Apple Intelligence centrally?

Apple gave no details on enterprise management at the announcement. Enterprise IT should record its requirements in advance and check the device management documentation once it appears. A policy should apply even before availability.

### Do companies need a policy for AI on smartphones?

Yes, because the features appear in everyday tools and employees often make data decisions unnoticed. A one-page policy with examples per data class is enough to start. It is revised after the pilot.

### How does Apple Intelligence differ from ChatGPT and Microsoft Copilot?

Apple Intelligence is integrated into the devices' operating system, ChatGPT is a separate service from OpenAI and Microsoft 365 Copilot works within the Office applications. Apple integrates ChatGPT as an optional partner. For IT these are three platforms with their own rules.

### When will Apple Intelligence be available in the UK and Europe?

Apple names autumn 2024 in US English as the first window, with further languages to follow over the course of the next year. Apple gave no date for British English or European languages at the announcement. Companies should start preparing regardless, because device inventory and policy need lead time.