The Biggest AI Paradox in Business: Employees Are Working Faster, but Leaders Aren’t Working Less

Businesses are no longer debating whether they need artificial intelligence. In 2026, the question has shifted: why are employees saving hours with AI while business leaders are not necessarily getting those hours back?

Recent research reveals a striking gap. According to McKinsey, 80% of respondents say AI has improved their individual productivity, yet only 37% report a positive impact on EBIT at the enterprise level.

Eagle Hill Consulting found a similar pattern in September 2026. AI is already being used in operations by 73% of surveyed organisations, in analytics and decision support by 72%, and in knowledge work and productivity by 71%. Meanwhile, 66% of executives report increased employee productivity and 59% report improved operational efficiency. Yet only 47% of organisations systematically identify opportunities to use AI across the business.

In other words, AI has learned how to make work faster. The next challenge is learning how to turn that speed into a fundamentally better way of working.

And nowhere is this gap more visible than in the relationship between a leader and their Executive Assistant.

Imagine that planning a business trip used to take an assistant two hours. With AI, it now takes twenty minutes.

At first glance, productivity has increased dramatically.

But there are two very different possible outcomes.

Scenario 1. The assistant spends twenty minutes finding ten hotels, several flight options and possible transfers, then sends everything to the executive with the question: “Which option would you like?”

Scenario 2. The assistant understands the executive’s schedule, travel preferences, budget parameters, meeting priorities and the boundaries of their own decision-making authority. AI speeds up the research, while the executive receives one recommended itinerary — or perhaps two options where their input is genuinely required.

The technology is the same.

The time saved by the assistant may be exactly the same.

But only in the second scenario has the executive’s attention been freed up as well.

That is why hours saved by employees are becoming an increasingly weak measure of AI effectiveness.

A much more useful question is:

How many tasks, approvals and minor decisions stopped reaching the executive altogether after AI was introduced?

On 21 September, IBM published a global study involving 1,500 CHROs and 8,800 employees.

80% of CHROs believe AI adoption is creating new forms of invisible work: employees need to review AI recommendations, correct errors, add context and handle exceptions. Meanwhile, 42% of employees say AI either increases their workload or creates additional work that often goes unnoticed.

This matters for any business investing in automation.

If a task used to take an hour and can now be completed in five minutes, that does not automatically mean the remaining 55 minutes have been converted into business value.

Someone still needs to determine:

✓ who reviews the AI-generated output;
✓ who is accountable when AI gets something wrong;
✓ which decisions must remain human-led;
✓ which exceptions need to be escalated;
✓ how employees should use the time that has been freed up.

Without this structure, a company may simply end up with much faster employees operating inside almost exactly the same organisational model.

И здесь появляется ещё один важный сдвиг.

По исследованию IBM, 71% CHRO называют способность контролировать, проверять и при необходимости отменять решения AI критически важным навыком. При этом только 29% сотрудников относят judgment — профессиональное суждение — к важным компетенциям.

Для рынка ассистентов это особенно показательно.

• Расшифровать встречу становится дешевле.
• Собрать информацию — быстрее.
• Сравнить варианты — проще.
• Подготовить первый черновик документа — тоже.

Но остаётся другая работа:

понять, что из полученного важно именно этому руководителю → добавить контекст → заметить исключение → оценить риск → принять решение в рамках полномочий → понять, нужно ли вообще подключать первое лицо.

Поэтому AI не обязательно уменьшает ценность сильного бизнес-ассистента.

Он может, наоборот, увеличивать разрыв между ассистентом-исполнителем и человеком, которому руководитель способен передать целый контур ответственности.

There is another important shift taking place.

According to IBM, 71% of CHROs say the ability to monitor, validate and override AI decisions is a critical skill. Yet only 29% of employees identify judgement as an important capability.

For the Executive Assistant market, this gap is particularly revealing.

Transcribing a meeting is becoming cheaper.

Researching information is becoming faster.

Comparing options is becoming easier.

Producing a first draft of a document is becoming almost instantaneous.

But another layer of work remains:

understanding what matters to this particular executive → adding context → spotting exceptions → assessing risk → making decisions within agreed boundaries → knowing whether the executive needs to be involved at all.

This is why AI does not necessarily make a strong Business Assistant less valuable.

It may actually widen the gap between an assistant who simply executes tasks and one who can take ownership of an entire area of responsibility.

Take something as simple as an AI-generated meeting summary.

AI produces a transcript and summary within minutes.

A basic-level assistant forwards the summary to the executive.

The executive still has to read it, identify the decisions made, work out who is responsible for what and decide what needs to happen next.

A higher-level assistant reviews the AI output, extracts the key agreements, assigns owners, sets follow-ups and escalates only those issues that genuinely require the executive’s decision.

AI worked in both cases.

But only in the second case did the architecture of the executive’s involvement actually change.

That is why, when hiring a Business Assistant today, it is no longer enough to ask which AI tools a candidate knows how to use.

A far more revealing question is:

What will you be able to hand over completely to this person now that technology can take care of part of the manual work?

Look at the last ten questions your assistant brought to you.

Divide them into four categories:

✓ my authority, strategic context or a high-stakes decision was genuinely required;
✓ the assistant simply did not have enough context;
✓ the boundaries of their authority had never been clearly defined;
✓ instead of bringing me a question, they could have brought me a proposed solution.

The final three categories represent potential executive attention that could be recovered.

This is a much more useful measure than counting prompts, AI tools or automatically generated documents.

The maturity of AI adoption can be measured by how much operational complexity stops travelling upwards through the organisation.

Eagle Hill’s research highlights another important issue: only 31% of surveyed organisations connect AI skills development with broader business and workforce planning.

The tool has been introduced. The employee is working faster. But their role, authority and underlying processes remain unchanged.

That is why the next stage of AI transformation is not another list of AI tools.

It is work redesign.

IBM found that organisations that clearly distinguish between human-led, AI-assisted and AI-executed processes report 18% lower risk and 20% higher quality.

For an executive and their assistant, this model can be reduced to three questions:

What can AI now do instead of the assistant?

What can the assistant fully take off the executive’s plate as a result?

Which decisions genuinely still need to reach the executive?

This is where businesses move beyond automating individual tasks and begin redesigning the system itself.

We regularly explore how founders and executives can reduce operational overload, build better delegation systems, work more effectively with Executive Assistants and use AI in a way that actually gives leaders their time back — rather than simply increasing the volume of completed tasks.

In our Telegram channel, Business on Autopilot, we share research, practical frameworks and real-world insights for entrepreneurs and senior leaders.

There is a simple way to tell.

Since introducing AI, has your assistant simply become faster at completing the same tasks — or have they started taking more operational complexity off your plate?

If it is the former, you have gained a more productive executor.

If it is the latter, the executive role itself begins to change.

That is why hiring an Executive Assistant in 2026 should no longer start with a simple list of duties. You need to understand which processes the person should own end to end, which decisions they can make independently, where the boundaries of their authority should sit and which competencies fit your particular leadership style.

At SMART AND TALENTED, you can book a free Executive Assistant profile assessment. We analyse the executive’s current workload, identify what can be delegated, define the level of autonomy required and build a profile of the right candidate.

Following the assessment, you will receive three relevant candidate CVs matched to the profile we have developed.