Why AI and automation projects overlook the admin that makes them work

The work that keeps an organisation running rarely gets named

When an organisation starts talking about AI and automation, the conversation usually begins with efficiency: faster decision-making, fewer manual steps, more scalable operations. What rarely gets named is the admin work that exists before any of that can happen.

Think about what happens upstream of any process you might want to automate. An approval workflow has to start somewhere. Data needs to be collected, organised and tagged before it can be processed by a system. A manager's inbox holds context that a chatbot cannot guess at. Invoices need to be routed to the right person, and that person needs to know what "right" even means in that organisation. None of this work is flashy. It does not show up in a transformation roadmap. It happens in the background, usually owned by administrators, PAs, operations staff or the person who just ended up holding the file.

This is the gap where AI and automation projects often fail. They aim high—imagine a world where this is automated—but they do not start low enough—here is what we actually do today, and why it matters.

Why this matters more than you might think

The projects that succeed at introducing AI and automation do so not because they had the cleverest technology, but because someone made the admin work visible first.

Imagine a charity that wants to automate initial case assessment. The leadership team agrees: we need to handle referrals faster, and a screening tool could flag urgent cases on arrival. Sensible idea. But the screening tool needs to know what "urgent" means for this charity. It needs to know how cases are currently triaged, what questions assessors ask, how they decide, and what information is already available in the system versus what is still on paper or in email threads. None of that is built into the software. It has to be extracted, documented and designed into the AI system before the tool can do anything useful.

The team that stops and names that work—interviews the staff who do it, maps the current process, writes down the rules and the exceptions—will implement something that works. The team that skips that step and just deploys the tool will launch something that produces either wrong answers or outputs that staff do not trust enough to act on.

The same pattern repeats across most domains: customer data consolidation (requires understanding what a "single customer" actually means in your business), automated reporting (depends on data being organised consistently in the first place), workflow automation (cannot route anything to anyone without knowing who the right person is and why), AI-assisted writing (needs brand guidelines and quality standards to be written down, not just held in someone's head).

What makes the admin visible, and how it changes the project

Making admin work visible does three things to an AI or automation project:

• It removes the fantasy that the system will "just work." Automation sits on top of information, processes and decisions. If those are disorganised, inconsistent or poorly understood, the automation amplifies the mess. A well-run admin practice creates the conditions where AI can actually work.
• It identifies what needs to happen before any technology is chosen. Often it is not more technology. It is clarity on how you currently work, documentation that does not exist yet, decisions that need to be made explicit, data that needs to be cleaned or retagged. That preparation work is less exciting than talking about the AI, but it is where most of the effort actually sits.
• It shifts the conversation from "can we automate this?" to "should we automate this, and what would it actually take?" Some things should not be automated yet, or ever. Some things are already handled well by a person, and adding automation creates friction rather than value. The admin-first view lets you choose deliberately instead of defaulting to "automate it because we can."

How to spot this gap in your own organisation

Here are some signals that admin is being overlooked in an AI or automation conversation:

• Technology is being chosen before the problem has been fully mapped. ("We are thinking about a no-code automation platform" before anyone has documented what the process actually is.)
• The people who do the work are not in the room where automation decisions are made. (This is usually because admin work is seen as separate from strategy, rather than foundational to it.)
• There is no clear answer to: "What information do we already have, where is it, and how is it currently organised?" Automation cannot work without knowing what it is working with.
• The success metric for the automation is just "it is deployed" or "we are using the tool." Using a tool is not the goal; doing better work with less repetition is. If the automation does not change that, it is not succeeding, even if it is technically running.
• No one has written down how decisions are currently made in this area. Automating a decision without naming the rules, exceptions and judgment calls involved usually ends badly.

What the preparation looks like

Before AI or automation enters the picture, the foundation is usually some combination of these:

• Process documentation: how is this work actually done today, step by step, including the edge cases and judgment calls that do not make it into job descriptions?
• Information audit: what data exists, where is it stored, in what state, who owns it, who needs access to it?
• Naming the decisions: where do humans currently make choices, and what information do they use to make them? This is what an automation might replicate or support.
• Identifying inconsistencies: where does the process vary from person to person, or day to day? Inconsistency in admin work is usually normal (context matters), but automation needs consistency to build on.
• Asking what is already working well: not everything needs to be redesigned. Some admin processes are fine as they are, and the value of an automation might be to support them better rather than replace them.

This preparation work is often done by the people who currently do the admin, working with someone who understands both the operational reality and the opportunity for AI and automation. It is not glamorous, but it is where the actual opportunity lives.

The people part

The other thing that usually gets overlooked is that introducing AI and automation changes what the people doing the work need to know and be able to do. An administrator who used to handle every referral assessment needs different skills if a screening system is handling the first pass and they are reviewing its outputs. They need to understand what the system is doing, how to spot when it is wrong, and what the override options are. That is training. That is change management. That is people, not just technology.

Organisations that succeed at AI and automation adoption invest in helping their team understand the new tools and the new expectations. That investment usually pays off in higher confidence, better quality control and faster adoption. Organisations that skip it end up with tools that staff do not trust, do not use properly, or actively work around.

Starting the conversation

If you are considering AI or automation for something in your organisation, here is a useful first question: "What admin work sits underneath this, and who does it today?" Not as a research project, but as part of understanding where the real opportunity sits and what would actually need to happen to make it work.

The organisations that get real value from AI and automation are not the ones chasing the technology. They are the ones that start by understanding their own everyday work: what keeps things running, where the repetition sits, where mistakes happen, what takes too long, what would free up time for higher-value work. That clarity is where the conversation should start.

If you want to explore where AI and automation could genuinely create value in your organisation, and what preparation would need to happen first, book a free Discovery Call. We will help you understand the admin and operational reality, spot the real opportunities, and decide whether AI and Automation Opportunity Mapping is the right next step.