AI at Work: Better Results, Same Broken Process

Two numbers from the same sample

The Eagle Hill Consulting survey, conducted by Ipsos among 200 US HR professionals, was released on 30 September 2026. It measures two facts inside the same group of people. 86% report better results thanks to artificial intelligence and automation.

51% spend at least half the working week^[1] on repetitive administrative tasks.

The distance between these two numbers is where the challenge described by this future-of-work research lives.

91% of respondents find information more accessible thanks to these tools. 70% say the time spent on routine work crowds out the questions that really matter. Both measures coexist in the same people, and together they describe a real gain resting on an unchanged workload.

The figure on the gain comes from the people who use the tools every day, so it holds as a measure of internal perception. The figure on the workload comes from the same sample, and that makes the comparison sturdier than any external estimate.

The same data, entered more than once

More than three quarters of respondents often enter the same data into different systems.

This is a structural signal. It describes an architecture where every tool asks again for what another tool already knows.

Automation applied to a path like this speeds up the automated stretch and leaves the joints between systems intact. The time saved inside one stage comes back as friction at the next handover. Measure the gain on a single task and you see an improvement; measure the whole week and you see the same week.

Double entry is also a data quality risk. Every manual copy opens the possibility of divergence between two records the board reads as a single source.

A function that re-keys data produces reports that need reconciliation before they reach the room.

A single point of data entry is a design choice, and it costs a governance decision more than a software licence. Organisations that take it cut repetitive work at the root. Those that postpone it buy tools that inherit the problem.

Trust runs faster than clarity

90% of respondents express trust in AI as support for their work, at least to some degree.

Alongside that trust, 37% report recurring integration difficulties. 36% are often or always in the dark about how an AI-generated result was reached. 32% regularly receive incomplete information.

Trust precedes the evidence, and that sequence deserves attention from the committee accountable for human capital. A function that relies on opaque output makes decisions about real people: hiring, promotions, development plans.

36% of declared opacity is a figure to carry to the board. It measures how much of the decision-making about people rests on a step the decision-maker would struggle to reconstruct.

The answer lives in process design: traceability of output, human review at the right point, written criteria. Functions that have already set these rules treat explainability as a procurement requirement, and they verify it before signing.

The bottleneck sits in work design

Melissa Jezior, president and chief executive officer of Eagle Hill Consulting, locates the HR function's challenge in how work is designed^[2] rather than in the technology stack.

That reading matches evidence already familiar to anyone following adoption. A tool dropped into an untouched process produces passive supervision, and the weekly hours of output checking measured in 2026 describe the same defect seen from another angle.

The organisation decides the larger share of the impact, the individual a smaller one. Adoption works as a designed condition rather than an individual choice.

Work design is a lever the HR function controls directly. Mapping the real path, stage by stage, costs weeks of analysis and opens the space where automation pays.

A redesign starts from the question of which stage generates the double entry, and proceeds by removing the step instead of accelerating it.

What the sample supports, and what stays open

Two hundred US HR professionals are a contained sample, and the survey should be read as a snapshot of one specific function.

The perimeter needs stating precisely: HR professionals, United States, self-reported data collected by Ipsos for Eagle Hill Consulting, published on 30 September 2026. Extending these percentages to everyone who works would require a different sample.

What the data does support is the internal comparison. The same 200 people report both the gain and the workload, so the distance between 86% and 51% lives inside a single experience.

Press coverage of the research picked up the point in the same week^[3], with attention to the gap between technology investment and time freed.

The objective measure stays open. A study based on observing the work, rather than on self-reporting, would say how much time really goes into the joints between systems. Until that measure exists, declared perception remains the best source available, and it should be read for what it is.

What the functions closing the gap do

Organisations that cut repetitive work act on three precise points, and they do so before buying the next tool.

• A single point of data entry, with automatic propagation to the other systems.
• A map of the real path before purchase, with a count of manual steps for each stage.
• An explainability requirement written into the contract, verified on real cases.

The first point attacks the double entry reported by more than three quarters of respondents. The second makes visible where time is lost, and moves the discussion from the vendor catalogue to the internal path.

The third point answers the 36% left in the dark about the criteria behind an output. A function that demands the explanation at procurement stage reaches decisions about people with full traceability.

Converting internal talent completes the picture. Training people who already know the path costs less than external recruitment, and it produces people able to redesign the work they practise every day.

This sequence asks for leadership time more than budget. The real cost sits in the weeks of attention a leader devotes to the map of the work, and that cost comes back as hours freed across the whole function.

The four conversations to take into the room

The evidence produces four distinct conversations, one for each role that decides.

• CEO: how much of the central functions' week still goes to repetitive tasks.
• CHRO: which stage of the HR path generates the double entry of data.
• CFO: what documented return process redesign has, compared with the additional licence.
• Talent and remuneration committee: what share of decisions about people rests on explainable output.

The CFO has the clearest case. The time freed by a single point of entry is measured in hours per person per week, and it compares against the cost of the redesign.

The talent committee has the newest metric. The share of decisions about people supported by documented criteria is a human capital indicator few boards monitor today.

Each role carries the same evidence through a different lens, and the coherence among the four readings decides the speed of the change. A board that hears four compatible versions of the same gap acts within a quarter. A board that hears four diverging diagnoses postpones.

The design question for CEO and CHRO reads like this: which stage of HR work deserves to be eliminated, rather than accelerated, at the next mapping of the path? The answer to that question is worth more than any new tool under evaluation.