Recruitment Automation for UK Agencies: What to Automate

Recruitment automation pays off when it takes the admin around people off consultants’ desks, and it goes wrong when it starts making decisions about people. For a UK agency that means automating scheduling, candidate and client updates, CRM data entry, compliance paperwork, timesheets and invoices, while shortlisting, rejection and offers stay with a consultant. Some of this is already inside the CRM you pay for, so check there first.

We build AI agents and automation that run inside the systems a business already uses. This article is written from that angle, so it says where a custom build earns its cost and where it does not. It covers what to automate, the UK GDPR and Equality Act 2010 rules on automated screening, and what to keep human. It also helps you choose between your CRM’s features, a bolt-on tool and a custom agent. It is general information and not legal advice.

Key takeaways

• Automate the admin around candidates and clients: scheduling, updates, CRM data entry, right-to-work paperwork, timesheets and invoices. Keep shortlisting, rejection and offers with a consultant.
• The UK GDPR rules on solely automated decisions were replaced in February 2026. In March 2026 the ICO reported that many employers using automated recruitment are likely relying on such decisions.
• Check your CRM’s own automation before buying or building. Bullhorn’s public product page, for example, describes self-booking interviews, stage-based messages and consent capture.
• A custom agent earns its cost when the work crosses systems and the hours saved beat the price. Ours start from £5,000.

How do you automate the recruitment process?

Start with the steps that follow fixed rules and repeat several times a week, then add AI only where the input is messy. Booking an interview follows rules and reading a CV is messy, while deciding whether to put someone forward is a judgement about a person that should stay with a consultant.

Recruitment task	Automate?	Why
Interview scheduling	Yes	Fixed rules, many repeats
Candidate and client updates	Yes, from templates	Same message, new data
CV parsing into the CRM	Yes, then check	Messy input, checkable output
Shortlisting and rejection	Assist only	A decision about a person
Right-to-work paperwork	Chase and file	A person or provider checks
Timesheets and invoices	Yes	Rules and arithmetic

The rest of this article takes each row in turn. Our guide to where automation actually pays off shows how to count the hours before you choose, and the workflow automation guide covers the general method.

Which recruitment tasks are worth automating first?

The ones a consultant does several times a day without thinking: booking, chasing and re-keying. Two groups are worth a closer look.

Scheduling and updates

Interview scheduling is the cleanest candidate for automation. The candidate picks from the consultant’s and the client’s free slots, the booking lands in both diaries, the CRM record updates and reminders go out. Candidate and client updates work the same way: a status change in the CRM triggers a templated message, so nobody has to chase to say there is no news yet.

Your CRM may already do this. Bullhorn’s public automation page, for instance, describes interviews that “book themselves” and tailored messages “at every stage of the hiring journey” (Bullhorn, retrieved October 2026). Other platforms may offer the same, so look at your CRM’s automation settings and marketplace before you buy anything.

CRM data: parsing, enrichment and hygiene

An agency runs on its data, and the data decays. Automation can parse an incoming CV into CRM fields and flag duplicate records. It can also fill in a client record from a public register and prompt for a refresh when a record goes stale. A consultant reviews anything the parser was unsure of, and a sensible set-up never overwrites a field a person has edited.

Re-engagement is the other big job: a rule that contacts candidates who have gone quiet. It needs a lawful basis and a retention rule. The ICO’s audit work on recruitment tools, covered next, raised concerns about candidate databases built without candidates’ knowledge and kept indefinitely.

CV screening: where AI helps and where it risks bias

AI can read a CV faster than a person, and the risk sits in what happens next. Used to summarise a CV, extract skills against the job spec and sort candidates into a list that a consultant reads, it saves time. Used to reject candidates automatically, it makes a decision about a person, and the law treats that differently.

In November 2024 the ICO published the outcome of its audits of AI recruitment tools. It made almost 300 recommendations, and reported that some tools were “allowing recruiters to filter out candidates with certain protected characteristics”. It was also concerned that some tools collected more personal information than necessary. These tools “retained it indefinitely to build large databases of potential candidates without their knowledge” (ICO, 6 November 2024). Its recommendations included giving candidates transparent privacy information with a clear retention period.

The government’s Department for Science, Innovation and Technology has also published a Responsible AI in Recruitment guide (25 March 2024). It focuses on “assurance good practice for the procurement and deployment of AI systems for HR and recruitment”. It is a sensible checklist to work through before you sign up to any screening tool.

Our design rule follows from this. The AI summarises and ranks, the consultant sees the CV and the reasons rather than a bare score, and only a person rejects a candidate. Every action is logged so you can show who decided what.

What do UK GDPR and the Equality Act say about automated decisions?

UK GDPR restricts significant decisions made solely by automated processing, and the wording changed in 2026. Section 80 of the Data (Use and Access) Act 2025 substituted a new set of rules, Articles 22A to 22D, for the old Article 22. The section’s commencement note gives 5 February 2026 as the date it came fully into force.

Under Article 22A a decision is “significant” if it produces a legal effect for the person or has “a similarly significant effect”. Where a significant decision is made solely by automated means, the safeguards in Article 22C include letting the person make representations, obtain human intervention and contest the decision.

The ICO has looked at recruitment specifically. On 31 March 2026 it published Recruitment rewired after speaking to more than 30 employers. Its conclusion was that “many employers engaging in automated recruitment are likely relying on solely automated decisions as part of this process”. The ICO adds that its findings “arise from voluntary engagement, not an audit or investigation”. It published draft guidance on automated decision-making the same day. At the time of writing (October 2026) that guidance is still marked as draft, so check the ICO site for the final version.

In plain terms, a solely automated decision is one with no meaningful human involvement. A consultant who only sees a score and clicks approve may not count, so the review step has to be real. The ICO’s advice for candidates sets the practical minimum for employers. Tell people when automated decisions are made about their application. Let them “contest the decision, explain your point of view and request a real person to take a second look” (ICO, 31 March 2026).

The Equality Act 2010 is the second risk. Section 19 covers indirect discrimination. It arises where a “provision, criterion or practice” applied to everyone puts people who share a protected characteristic at a “particular disadvantage”. The employer then has to show that it is “a proportionate means of achieving a legitimate aim”. A rule written into a screening tool is easy to treat as a technical setting. It is worth asking whether it is a practice of this kind.

The Home Office’s right-to-work guide notes that a discrimination complaint can be brought against “an employer, a prospective employer or an employment agency” (GOV.UK, 1 October 2026 edition). Agencies are not on the sidelines. If you plan to screen with AI, take legal advice on your set-up first.

What should stay with a consultant?

Anything where the output is a judgement about a person, or a relationship that you want to keep. Automation can prepare the ground for each of these, and a person makes the call.

• Shortlisting and rejection. The tool can summarise and rank. The consultant decides who goes forward and who does not.
• Client briefing and advice. The CRM history can be pulled together for the call, but the conversation is yours.
• Offers, negotiation and difficult feedback. These are the moments a candidate or client remembers.
• Challenges to a decision. When a candidate says the tool got it wrong, the request must reach a person quickly. The ICO says candidates are entitled to that.

Right-to-work checks, timesheets and invoices

These are the compliance and finance jobs where automation does the most routine work with the least risk, as long as you are clear about who is responsible.

Right-to-work checks. The Home Office’s employer guide (1 October 2026 edition) puts the duty on the employer. It says: “With the exception of prescribed checks carried out by a RtW DVSP, the employer is responsible for carrying out the check” (GOV.UK). A RtW DVSP is a registered right-to-work digital verification service provider. For temporary workers, it says the employment business that directly engages agency workers and supplies them to a hirer is responsible for the check. Records must be kept “for the duration of the worker’s employment and for two years afterwards”. The guide also lists a civil penalty of up to £60,000 per illegal worker where the prescribed checks have not been carried out.

So automation chases the worker for documents or a share code and books the check. It files the dated record with the right retention date and alerts the consultant when something is missing. The check itself follows the Home Office’s prescribed methods. Check the current guide, because it changes.

Timesheets and invoices. An approval in one system should create the invoice in another without anyone retyping it. We built that pattern for an international events group, which is not a recruiter. We connected its CRM (Pipedrive), contract tool (SmartDocs) and Xero without replacing them. When a contract is signed the CRM updates, and the invoice is raised in Xero automatically, as the events group sales automation case study describes. Around 500 sales invoices a month no longer get typed by hand. A placement fee or an approved timesheet follows the same pattern, and it is the kind of job systems integration covers, with two systems connected from £5,000.

Switch on, buy or build?

Start with what you already pay for. If your CRM or ATS ships the feature, switching it on costs the least and adds no new system to maintain. The three routes compare like this.

Route	Best when	Watch for
Switch on your CRM’s automation	Simple rules, one system	Limits on what it can touch
Buy an add-on tool	A standard job the CRM lacks	Another login, another copy of data
Build an agent in your CRM	Work crosses systems or uses your rules	Needs careful scoping first

A custom agent beats a bolt-on tool in the gaps between systems. Examples include a screening summary written into the right CRM fields in your format, and re-engagement driven by your own placement and consent data. Others are record cleaning across the ATS, job boards and mailbox, or timesheet-to-invoice across your finance package. If “agent” is a new term, our explainer on what an AI agent is is the place to start.

Our AI agents service has two published tiers. A first agent (one job, scoped) is from £5,000, and an agent rollout across several workflows is from £12,000. A first agent works inside your existing tools, with clear limits on what it decides alone. For broader work across your processes, see process automation.

We built a similar pattern for an energy broker. Replies are classified and responses drafted for approval, and an 85% confidence threshold decides what runs automatically and what waits for a person (SBU). A candidate inbox suits the same approach: routine questions answered, and anything uncertain or sensitive passed to a consultant.

A worked example (illustrative)

Say three consultants each spend 4 hours a week re-keying CVs and chasing candidates, over 46 working weeks, and an hour of consultant time costs the agency £30. That is 3 × 4 × 46 = 552 hours, or £16,560 a year. If an agent removes half of that, the time saved is worth £8,280 a year. Set that against a first-agent price from £5,000, plus running costs such as hosting, model usage and, if you take it, maintenance and support from £400 a month. If your own figures come out at a tenth of this example, a build will not pay back and your CRM’s own feature is the better answer.

As a rule of thumb, smaller agencies usually get more from their CRM’s tools and an ATS add-on than from a custom build. A build starts to make sense when several consultants lose hours to the same cross-system job.

A first step that costs little

Pick one workflow, count the hours it takes and check the data protection position before you automate it.

1. Choose a rule-based job. Scheduling, status updates and invoice creation are the safest starts.
2. Check your CRM first. Look at its automation settings and marketplace for the same job.
3. Write down where a person decides. Add a notice to candidates, a route to a human and a retention period for the data.
4. Pilot with one team. Compare hours before and after, then extend.

If the job crosses systems, scope a first AI agent with us. It covers one named job inside your CRM, with sign-off where it matters. Scope and price are fixed before work starts and you pay by milestone as each stage is delivered. You can cancel within the first 21 days of a build for a full refund. If the main need is joining your CRM to your finance or document tools, systems integration is the closer fit.