In 2006, Anikka Fragodt applied for the position of Facebook CEO Mark Zuckerberg’s personal assistant after seeing the listing on a job board. In the book the Facebook Effect, author David Kirkpatrick describes Fragodt as Zuckerberg’s ‘trusted personal assistant’. Her long list of responsibilities kept him on track; she was the only person responsible for his calendar; she organised his travel arrangements and planned Facebook’s global conferences and events. She was also entrusted with planning his wedding in 2012, which was done in secret.

Fragodt, who served as Zuckerberg’s assistant from 2006–2013, became known as his ‘protector’ and was often referred to as the most influential woman at Facebook. According to the NZ Herald, Zuckerberg credited her with much of his success, saying, ‘Annika helped me become a better CEO,’ and acknowledged her role in keeping the operations running smoothly while he focused on expanding the company.

Now the Meta founder and CEO, who the Financial Review describes as being ‘consumed by his AI [artificial intelligence] ambitions’ has no need to advertise for an assistant. Instead, as reported by the Wall Street Journal in March this year, he is developing a personal AI agent to help him with running the company. The initiative is part of a broader push to equip every Meta employee with their own AI assistant.

The system, which is still in development and training stages, will focus on accelerating access to information. It can provide answers that would otherwise require going through layers of management and personnel, thereby avoiding delays that are standard in many large organisations. According to the report, the project ‘reflects a wider effort to speed up decision-making, flatten organisational structures and reshape day-to-day work at Meta, as the company seeks to remain competitive with fast-moving AI start-ups’.

Company insiders describe Zuckerberg’s personal agent as ‘a co-pilot that augments human decision-makers’. According to the Financial Express, the agent is expected to be more than an internal search engine. It is designed to understand context, combine multiple pieces of information across projects and surface options or trade-offs when it puts an answer together.

During the company’s fourth-quarter earnings call in January, Zuckerberg declared that this year would be when ‘AI starts to dramatically change the way’ Meta works. ‘We’re investing in AI-native tooling so individuals at Meta can get more done. We’re elevating individual contributors and flattening teams,’ he said. ‘If we do this, then I think that we’re going to get a lot more done, and I think it’ll be a lot more fun.’ He has also been spending more time coding as these systems evolve. The project is intended to help him obtain information more efficiently, removing the need to go through multiple people to get the same answers.

While this may sound like a dream scenario, an article on the Next Web explains that building an AI agent capable of ‘assisting a chief executive is a different challenge from deploying coding assistants or general information tools. A system designed to help run a company must navigate competing priorities, weigh strategic decisions with incomplete information, and develop a contextual understanding of organisational politics and human relationships. These tasks strain even sophisticated AI systems’.

The challenges are not only operational, though. As AI systems become more involved in decision-making at an upper management level, questions around governance and accountability are also coming into focus. ‘A key concern is the emergence of accountability gaps when AI systems perform or influence executive-level tasks,’ says Lucia Quy, an associate director at Webber Wentzel Fusion, the legal firm’s AI advisory subsidiary.

‘If an AI bot acts unethically, unlawfully, or causes harm, liability may be difficult to allocate between the developer, deployer, organisation and individual user. South Africa’s evolving AI governance approach emphasises human oversight, accountability, explainability, fairness and board-level responsibility, reflecting the principle that organisations cannot outsource legal accountability to AI systems,’ she says.

Quy argues that when it comes to legislation and a national AI policy in SA, stronger governance detail is required.

‘The policy should provide clearer mandates, powers and accountability mechanisms for oversight bodies, together with sector-specific guidance that translates principles such as “sufficient explainability” into practical business requirements,’ she says.

By automating the layer that would ordinarily require hours of human attention, Zuckerberg is claiming back time for the decisions that machines cannot yet reasonably make.

The tech giant may be pushing the concept furthest, but Zuckerberg is far from the only chief executive relying on AI in his daily work. In an interview with the Wall Street Journal, Apple CEO Tim Cook said Apple Intelligence helps him summarise long emails. ‘If I can save time here and there, it adds up to something significant across a day, a week, a month. It’s changed my life. It really has,’ he said.

That seems to be one of the main reasons CEOs are turning to AI apps – to save time and therefore become more productive. But can AI replace personal and executive assistants, and would the big bosses want it to? Last year, Medium reviewed surveys, research and feedback from 500 CEOs and founders across industries. The online publishing platform and social network says their answers revealed something clear: AI can help you move faster, but it can’t help you lead better.

When asked what they value most about their executive assistants, CEOs didn’t say ‘calendar scheduling’ or ‘travel bookings’.

The top three responses were: judgement and decision support; emotional intelligence; and anticipation. ‘My assistant knows what I need before I do,’ said the CEO of a fintech firm. ‘No algorithm does that.’ They also feel that AI still struggles with human nuance, context awareness, ethical judgement and trust, in that you can’t ask it to represent your values in sensitive conversations. A tech founder explained further. ‘I trust my assistant with my inbox because I trust her judgement. I can’t trust a bot not to send the wrong email at the wrong time.’

Medium says the point is subtle but essential: AI can organise, but it can’t understand context. The article goes on to say that while AI has limitations, it has become essential to CEOs’ work, as well as their assistants’ daily responsibilities. It quotes a McKinsey report that shows that executives using AI to automate admin work save an average of 12 hours a week, the equivalent of 600 hours per year.

It could prove to be a powerful combination, with AI helping with the heavy admin lifting and the human assistant adding the personal touches. Senior executives don’t seem to be in a rush to replace their trusted assistants either. Among the CEOs surveyed, 82% use at least one AI productivity tool daily; 63% believe AI has improved the efficiency of their executive assistant relationship and 18% have replaced fundamental admin roles with AI, primarily for repetitive data tasks. But only 7% said they could see AI replacing a strategic executive assistant in the next 10 years. One of the main reasons for this was that they felt great assistants don’t just react, they think ahead.

So it seems that the often-indispensable assistants are still just that – AI isn’t replacing them, but it is removing much of their, and business leaders’, administrative burdens. It isn’t a case of human versus machine, but rather human and machine working together. As one Fortune 500 CEO put it: ‘My assistant with AI is unstoppable. My assistant without it would drown.’

By Philippa Byron
Image: iStock