AI in recruitment: the tech stack is full, so why are recruiters still doing so much admin?

There is an ATS. A CRM. LinkedIn. Job boards. Email. Calendars. Phone systems. Messaging platforms. Video interviewing. CV parsing. Automation tools. Reporting platforms.

And, somewhere in the middle of all that, probably a spreadsheet doing something more important than anyone would like to admit.

So when the answer to recruitment productivity is presented as another AI tool, I’m not convinced.

The problem isn’t necessarily that agencies need more software.

Quite often, they need the software they already have to work together properly.

That’s where I think the current shift towards AI agents becomes much more interesting.

Not because we suddenly need an artificial recruiter.

Because recruitment is full of small gaps between systems, people and actions. Consultants spend an extraordinary amount of time bridging those gaps manually.

The recruitment stack has grown. The workflow hasn’t always caught up.

Consider a fairly ordinary candidate journey.

1. Someone applies for a role.
2. Their CV enters the ATS.
3. A recruiter reviews it.
4. Perhaps some information is missing, so they send an email.
5. The candidate replies.
6. Someone updates the CRM.
7. A screening call needs arranging.
8. Calendars are checked.
9. A call takes place.
10. Notes are written.
11. The candidate is shortlisted.
12. The client wants an interview.
13. More diary coordination follows.
14. Then come confirmations, reminders, feedback, another interview, another update and eventually an offer or rejection.

None of those steps is particularly unusual.

The problem is how many times a person has to notice something, move information somewhere else or remember what needs to happen next.

That’s not really a recruitment problem.

It’s a systems problem.

AI in recruitment has already proved it can save time

There is enough evidence around AI in recruitment now to move beyond speculation.

LinkedIn’s Future of Recruiting research found that talent acquisition professionals already using generative AI reported an average 20% reduction in their workload. That’s roughly one working day per week. LinkedIn also found that employers were dramatically increasing their emphasis on relationship development as a recruiter skill.

Bullhorn’s 2026 UK and Ireland research tells a similar story from the agency side.

Among nearly 300 UK&I recruitment professionals surveyed, agencies using AI at some point in the recruitment process were four to eight times more likely to have grown revenue than lost revenue in 2025. Bullhorn also found that 26% of agencies had moved beyond basic generative AI into some level of agentic AI.

That does not prove AI caused the revenue growth. Better-run agencies may also be more likely to invest in technology in the first place.

But the operational findings are harder to ignore.

Recruiters told Bullhorn that searching for candidates takes up more of their time than anything else, followed by screening. Most said AI had reduced the time spent on search and screening by somewhere between 26% and 75%.

So the efficiency case is becoming fairly clear.

The next question is what we do with it.

Another AI window isn’t the answer

We’ve all seen the basic implementation. A recruiter opens the ATS. Then opens ChatGPT. Copies some information across. Writes a prompt. Copies the result back. Perhaps opens another system to send it.

Technically, AI has been introduced.

Operationally, we’ve just added another tab.

That isn’t transformation. The more useful idea is to make AI part of the workflow itself.

1. A candidate applies.
2. The system identifies what information is already available.
3. It recognises what’s missing.
4. It contacts the candidate if appropriate.
5. The response goes back into the ATS.
6. If the candidate meets the agreed criteria for a registration conversation, the system checks the consultant’s availability and offers an appointment.
7. The booking is made.
8. Confirmation goes out.
9. The recruiter arrives at the call with the information already organised.

Nobody had to copy and paste anything between three systems.

That’s what we mean when we talk about business automation at Megabite. The objective isn’t to add AI for the sake of saying the agency uses AI. It’s to remove unnecessary movement between one part of the process and the next.

Recruitment is particularly suited to this

Recruitment contains a useful mixture of two very different types of work.

The Recruitment & Employment Confederation put it neatly this summer.

I think that’s exactly right.

Think about the ATS as the source of truth, not the whole solution

Most established agencies don’t need somebody turning up and telling them to throw away their ATS.

The ATS is usually where the history lives. Candidates. Clients. Jobs. Placements. Notes. Activity. Years of data that has commercial value precisely because the agency has spent years creating it.

The challenge is getting more value from that information without asking consultants to become database administrators.

This is where integration matters. A good AI or automation layer should be able to work with the systems the agency already trusts.

• The ATS knows a contractor’s assignment ends in three weeks. The workflow recognises that event, checks whether the consultant has already started the redeployment conversation, and creates the action or prepares the outreach if not.
• A new vacancy arrives requiring skills held by candidates already in the database. Instead of relying on somebody remembering those candidates or constructing the perfect search, the system can surface them.
• A client enquiry lands after office hours. Instead of becoming tomorrow morning’s unread email, it can be acknowledged, qualified and routed.

This is where systems integration becomes much more important than simply buying another AI product.

The value often sits between the systems.

The inbox is still doing far too much work

This is one of the things we look for when reviewing business processes. How much operational activity is effectively being managed through email?

A new requirement comes in by email. A CV arrives. A client replies. An interview time changes. A candidate sends a document. Someone needs to notice it and translate that email into an action somewhere else.

Email is excellent communication technology. It is not a particularly good workflow engine.

The same applies to phone calls. If a hiring manager rings about a new vacancy and the consultant is unavailable, what happens? A message? An email notification? A note to call back? Or can the enquiry be captured properly, the basic requirement understood, the CRM updated and a conversation booked?

A good lead response process should reduce the amount of commercial intent that gets turned into internal administration.

That’s particularly important in recruitment because the agency has two customer journeys running at the same time: candidates and clients.

Both generate communication. Both generate data. Both create follow-up. And both suffer when information stops moving.

Search is an obvious place for AI. Decision-making needs more care.

There’s a temptation to jump from AI can find relevant candidates to AI should decide which candidates progress. Those are not the same thing.

AI search can be extremely useful because recruitment databases are messy. Job titles vary. Candidates describe the same skills differently. Good people disappear into old records. Keyword searches are limited by the words somebody thinks to type.

Using AI to surface potential matches makes sense. Using an opaque score to automatically reject someone deserves much more scrutiny.

The Information Commissioner’s Office has made recruitment a specific regulatory focus. Its 2024 audits of AI recruitment providers resulted in almost 300 recommendations, covering issues including excessive collection of personal information, transparency, fairness and the use of protected characteristics.

In 2026 the ICO then examined automated recruitment practices across more than 30 employers. It concluded that many appeared to be using solely automated decisions with potentially significant effects on candidates, which can bring additional UK GDPR safeguards into play.

That’s an important distinction when deciding where AI in recruitment should actually be used.

Scheduling an interview is one thing. Automatically deciding somebody isn’t worthy of one is another.

The best automation is often surprisingly boring

This is something we’ve found repeatedly when looking at business automation. The glamorous demo is rarely where the biggest operational gain sits.

It’s often things such as:

• A candidate registration being completed properly before the consultant calls
• An interview being arranged without six emails
• A consultant being reminded about a contractor approaching the end of an assignment
• A dormant candidate being surfaced when a relevant job appears
• A client enquiry being qualified outside office hours
• A missing document being chased automatically
• A CRM record being updated after a conversation instead of at the end of the day

Individually, none of these is revolutionary. Together, they change how much of a recruiter’s day is spent moving work around.

And that matters in the current market. The REC says the UK recruitment industry contributed £40.6 billion to the economy in 2024, but the sector has been operating through weaker demand, regulatory change and considerable economic uncertainty.

Productivity isn’t an abstract technology goal in that environment. It’s commercial.

Where we start with a recruitment agency

Not with a list of AI products. We start by watching how work actually moves.

• Where does a client enquiry enter the business, and what happens next?
• Where does candidate information get duplicated?
• Which activities require someone to manually move data between the ATS, inbox, calendar and phone?
• What gets forgotten when consultants get busy?
• What do the best billers do manually that could be systemised for everyone?
• Which processes genuinely require recruiter judgement?
• And which ones simply require something to happen reliably?

Then we pick one. Not ten. One process with a measurable outcome.

It might be faster response to new client enquiries. It might be candidate registration. It might be interview coordination. It might be redeployment. It might be dormant candidate activation.

Then we map the current workflow, remove unnecessary steps and decide where automation or AI genuinely helps.

That’s how we approach automation at Megabite. We’re not interested in replacing a capable ATS simply because somebody has launched another AI platform.

Recruitment agencies already have plenty of technology. The opportunity is often in connecting what is already there and removing the manual gaps between CRM, telephone, email, calendars and people.

Sometimes the answer is AI. Sometimes it’s straightforward automation. Sometimes it’s better integration. And sometimes the existing human process is perfectly good and should be left alone.

The goal isn’t fewer recruiters

It’s recruiters doing more recruitment.

The interesting thing about AI in recruitment is that the more capable the technology becomes, the clearer the human part of the job becomes as well.

LinkedIn’s research points towards the same conclusion: AI is increasingly taking on routine work while recruiters are being pushed towards consulting, judgement, relationship-building and deeper candidate engagement.

Those are the parts worth protecting. A good recruiter knows when a client’s brief doesn’t make sense. They know how to get somebody interested in a role they weren’t looking for. They can read uncertainty in a conversation. They can challenge a hiring manager. They can rescue an offer that’s starting to fall apart.

No ATS integration replaces that.

But there is no great human value in spending fifteen minutes finding an interview time.

That’s why I don’t think UK recruitment needs another wave of disconnected AI tools. It needs fewer disconnected processes.

Get the ATS, inbox, phone, calendar and automation layer working together. Let software deal with the predictable movement of information. Let recruiters deal with people.

That’s a much more interesting use of AI. And probably a much more profitable one too.

Common questions

Will AI replace recruiters?

No, and that is not really the interesting question. AI is well suited to the operational half of recruitment — searching, scheduling, chasing information, updating records — and poorly suited to the human half. Understanding what a client actually needs, persuading a good candidate to consider something they had not planned, reading uncertainty in a conversation and rescuing an offer that is falling apart are the parts worth protecting. The goal is not fewer recruiters. It is recruiters doing more recruitment.

Do we need to replace our ATS to use AI in recruitment?

Usually not. The ATS is normally where the history lives — candidates, clients, jobs, placements, notes and activity — and that data has commercial value precisely because the agency spent years creating it. The more useful question is whether an AI or automation layer can work with the systems the agency already trusts. In recruitment the value tends to sit between the systems rather than inside any one of them.

How much time does AI actually save recruiters?

LinkedIn’s Future of Recruiting research found that talent professionals using generative AI reported an average 20% reduction in workload, roughly one working day a week. Bullhorn’s 2026 UK and Ireland research found most recruiters said AI had cut the time spent on search and screening by between 26% and 75%. Both figures are self-reported, so treat them as an indication of direction rather than a guaranteed return.

Which recruitment process should an agency automate first?

One, not ten, and one with a measurable outcome. In practice that is usually client enquiry response, candidate registration, interview coordination, redeployment of contractors approaching the end of an assignment, or dormant candidate activation. Which one depends on where information currently stops moving, which is why the work starts by mapping the existing workflow rather than by choosing a product.

Can AI reject candidates automatically?

Technically yes. Legally it is a different matter. The ICO examined automated recruitment practices across more than 30 employers in 2026 and concluded that many appeared to be making solely automated decisions with potentially significant effects on candidates, which can bring additional UK GDPR safeguards into play. Its 2024 audits of AI recruitment providers produced almost 300 recommendations. Using AI to surface candidates from a messy database is a reasonable use. Using an opaque score to reject someone deserves much more scrutiny.

What does AI in recruitment look like day to day?

Mostly unglamorous. A candidate registration completed properly before the consultant calls. An interview arranged without six emails. A reminder about a contractor whose assignment ends in three weeks. A dormant candidate surfaced when a relevant job appears. A client enquiry qualified outside office hours. A CRM record updated after a conversation rather than at the end of the day. Individually none of these is revolutionary. Together they change how much of a recruiter’s day is spent moving work around.