FeaturesAIAI Department Managers

An AI manager for every department - on the data you already capture.

Seven specialist AI managers across Health & Safety, Attendance, Checklists, Scheduling, Training, HR and Recruitment - brought together by a single daily executive briefing across the entire operation. We haven't found another UK workforce platform that ships every module and a specialist AI manager for each one in a single suite. The result - your General Managers and Duty Managers stop chasing punctuality, Bradford scores, overdue checklists, expired training certificates and risk assessments due for annual review, because the AI managers surface every one of those daily and route the manager straight to the action. Built UK-first - advisory only, PII anonymised before any model call, GDPR Article 22 safe. No autonomous decisions. No surprises.

Daily executive summary
Cross-module synthesis · refreshes every 5 minutes · Generated at 08:00
AIConfidence: 92%

Across your seven managers this morning: one RIDDOR is unreported at Manchester (Crown), and an accident pattern is forming on Friday closing shifts. Punctuality is drifting in the late kitchen shift across Camden and Shoreditch — three clock-ins past 18:08 in the last five days.

Six training certificates expire in the next 14 days; the Food Safety Level 2 lapse at Leeds is most urgent. Two new Bradford movements (Jamie K. and Hana T.) suggest wellbeing conversations — not formal action.

Health & Safety
RIDDOR unreported
4 openView
Attendance
Punctuality drift
3 openView
Checklists
1 missed last night
2 openView
Training
Certs expiring 14d
6 openView
HR
Wellbeing checks
2 openView
Recruitment
47 candidates
4 openView
AI Health & Safety Manager
Briefing · today 08:00 · scoped to The Crown, Manchester
AIAdvisory
  • Unreported RIDDOR · slip incident, kitchen
    14 Mar · 4 hours since shift handover · regulatory deadline approaching
  • Working at Height RA — annual review
    Last reviewed 11 Mar 2024 · 6 days overdue
  • 3 COSHH assessments due-soon
    Caustic cleaner · Bleach · Sanitiser
  • Accident pattern · Friday close 22:30–23:00
    Wet floor cluster · 3 incidents in 6 weeks · controls updated
    Acknowledged by James Okafor · 08:42
Advisory only · GDPR Article 22 audit trail
Signals that should reach a manager - and don't

Seven things UK operators learn about too late.

Either the data exists somewhere and nobody surfaces it, or the system to capture it was never in place - Bradford scoring, Risk Assessment Reviews, training-certificate countdowns aren't standard kit in most operations. Either way the manager finds out from the wrong source - an EHO, a tribunal letter, a Sunday spreadsheet - when a daily brief should have arrived first thing. AssistantManagerHQ is one platform across HR, attendance, checklists, training, recruitment and H&S - and seven specialist AI managers that capture what's missing, surface what's already there, and brief the manager who can act.

Your General Managers and Duty Managers spend half their week chasing the same five things - late clock-ins, Bradford-style absence drift, missed checklists, expiring training certificates and Risk Assessment Reviews due - because nothing surfaces them automatically.

Cost:A manager who should be running the shift is running a spreadsheet - and the items they're chasing are exactly the ones an EHO inspection, an HSE audit or a tribunal will ask for first.

Fixed:The seven AI Department Managers surface every one of those items on a single daily brief, route the manager straight to the action with one click, and brief the executive summary so the operations director doesn't have to ask. The General Manager and Duty Manager stop chasing and start running - and the platform proves the chase happened, with a timestamp and a name.

Most operations don't run Bradford-style absence scoring at all - the pattern only becomes visible to the manager when it's already a performance or wellbeing problem.

Cost:A wellbeing conversation that should have happened in week six happens in week sixteen - by which point it's a disciplinary, not a check-in. Or the manager opens a disciplinary on a pattern that turns out to be a childcare clash.

Fixed:AssistantManagerHQ captures absence into the Bradford Factor module by code, with a wellbeing-first gate that blocks any disciplinary route until a wellbeing check is recorded. The AI HR Manager surfaces patterns as wellbeing-conversation starters - pseudonymised before any model call, signed off by a named human.

A late clock-in is captured by the time clock and nothing else happens - HR records don't update, no absence trail forms, the next 1-to-1 doesn't mention it.

Cost:Punctuality drift that's been visible to the time clock for three months only becomes a conversation when it's already a performance problem - by which point there's no early record of when the manager first noticed.

Fixed:AssistantManagerHQ ties the Time Clock to HR records, and the AI Attendance Manager surfaces punctuality drift, late-arrival clustering and absence patterns on the daily brief - one click into the wellbeing or performance conversation, with a timestamped trail.

A Working at Height Risk Assessment passed its annual review date six months ago - the document exists in a folder somewhere; the renewal trigger doesn't exist anywhere.

Cost:An HSE inspector arrives and the conversation starts with a six-month-old date stamp. There's no defensible answer to "when did you know it was overdue?"

Fixed:AssistantManagerHQ captures Risk Assessments as live records with review-due dates, and the AI Health & Safety Manager surfaces every assessment past its review date on the daily brief - one click into the document, the assessor rates new control measures, the audit trail names the human and the time.

Three Food Safety Level 2 certificates expire in 14 days - the training records may exist in a spreadsheet, an LMS, a paper folder or nowhere at all, and the site manager doesn't find out until one of them lapses.

Cost:An EHO visit lands the week after a certificate expires, and the conversation is whether you knew. "We didn't see it coming" is not a defensible answer.

Fixed:AssistantManagerHQ holds training records on the employee record with expiry dates, and the AI Training Manager surfaces every certificate expiring within 14 / 30 / 60 days - broken down by role and site - on the daily brief.

AI vendors promise autonomous workflows - and UK operators end up at a tribunal because something fired before anyone noticed.

Cost:A candidate auto-rejected with their name still on the CV, a disciplinary triggered by an absence score, a rota published over an unavailable employee. The first time anyone notices is the tribunal letter.

Fixed:Every AI output across AssistantManagerHQ is advisory only - a draft, a brief, a ranking, a nudge - that a named human signs off. No code path lets AI submit a disciplinary, dismiss a candidate, publish a rota or send an external communication on its own. GDPR Article 22 by design.

Eighty-seven candidates apply for twelve night-cleaning roles - they land in an inbox or an ATS in three different formats, and the hiring manager hasn't opened any of them by Friday.

Cost:By Monday the strongest candidates have accepted offers elsewhere; the ones you eventually shortlist are the ones who waited. The decision trail is "the ones still available."

Fixed:The AI Recruitment Manager ranks every applicant on an anonymised CV against role criteria within minutes - names, postcodes, DOB and NI numbers stripped before the model call. The hiring manager opens the top 20 instead of all 87, and the shortlist decision sits on the candidate record with a named human and a timestamp. Equality Act 2010 + GDPR Article 22 by design.

7 specialist
AI managers across the suite
Advisory only
GDPR Article 22 safe
PII anonymised
before any model call
Named human
signs off every decision
The promise

One AI suite that knits the whole operation together - and frees your managers.

AssistantManagerHQ AI is built on three rules. One - AI reads the data you already capture, not a generic dataset; rotas, clock-ins, checklists, candidates, absences, accidents are what the model sees. Two - every output is advisory; a brief, a draft, a ranking, a nudge - never a decision. The hiring manager shortlists, the safety manager files the RIDDOR, the rota manager publishes the schedule, the HR manager opens the disciplinary. Three - UK regulation is in the design; GDPR Article 22 is honoured by code (no autonomous decisions affecting individuals), and PII is anonymised before any candidate or absence call. Where AssistantManagerHQ is genuinely different - seven specialist AI managers across every operational area in one suite, brought together by a single cross-module executive summary that synthesises the day across Health & Safety, Attendance, Checklists, Scheduling, Training, HR and Recruitment. We haven't found another UK workforce platform that does this - most ship one or two AI assistants bolted onto one or two modules, not seven specialists with a unified brief. The operational result - your General Managers and Duty Managers stop spending their week chasing punctuality, Bradford scores, overdue checklists, expired training certificates and risk assessments due for annual review; the AI managers surface every one of those daily and route the manager straight to the fix. Across the rest of the platform, AI drafts the documents - risk-assessment control measures, COSHH substance identifications, job descriptions, interview scorecards, training questions, draft rotas - that a human edits and signs off.

READ
Operational data sources
  • Published rotas
  • Clock-in records
  • Checklist runs
  • Risk assessments
  • COSHH assessments
  • Accident records
  • Candidate applications
  • Absence spells
  • Training courses
No external dataset · no scraping · no third-party feeds
BRIEF
H&S ManagerAI
RIDDOR unreported08:00
RA due-soon (1)08:00
COSHH due-soon (3)08:00
Pattern detected · Fri close08:00
+ 6 other managers brief on their own data
SIGN OFF
Human acknowledgement
Sarah Khan · Operations Lead
Acknowledged · Friday pattern
James Okafor · H&S Lead
Opened wellbeing check (not disciplinary)
Naomi Adeyemi · HR Manager
Shortlisted 12 candidates · rejected 4
GDPR Article 22 · named human · timestamped
  • Seven specialist AI managers - one for every operational area, briefing on real data.
  • Advisory only - no autonomous disciplinaries, no autonomous hiring, no autonomous rota publishing.
  • PII anonymised - names, emails, postcodes, DOB stripped from candidate CVs and absence insights before the model call.
  • GDPR Article 22 by design - every decision affecting an individual is signed off by a named human with a timestamp.
  • Built on OpenAI GPT-4 family + GPT-4 Vision + Azure Document Intelligence for OCR.
  • Drafts that a human signs off - risk assessments, COSHH, job descriptions, scorecards, training questions, rotas.
How it works

From the data you already capture to the draft you sign off - automatically.

STEP 01

Read

Each AI manager reads only the data your business already captures - rotas, clock-ins, checklists, accidents, candidates, absences, courses. No external dataset, no scraping, no third-party feeds.

STEP 02

Brief

A daily briefing per manager surfaces what's overdue, what's drifting, what needs a conversation - with action buttons that link straight to the module, never an auto-action.

STEP 03

Draft

Where a document is needed, AI drafts it - risk-assessment control measures, COSHH substance identification from a phone photo, a job description, an interview scorecard, training questions, a rota - and presents it for review.

STEP 04

Sign off

Every AI output requires human acknowledgement before any downstream effect - the manager rates and saves the RA, confirms the COSHH substance, edits and posts the job description, approves and publishes the rota. The audit trail names the human and the time.

The status quo vs. AssistantManagerHQ

Why a unified AI suite beats five AI assistants in five tabs.

Five tabs, autonomous decisions, a manager's Sunday on a spreadsheet
  • Punctuality, Bradford, checklists, training certs and RA renewals chased manually across five systems.
  • AI bolted onto one or two modules; nothing knits the operation together.
  • Autonomous candidate rejection on a CV that included the name and postcode.
  • Disciplinary opened because a model flagged a pattern.
  • Rotas published over unavailable employees because the AI "optimised".
AssistantManagerHQ AI Department Managers
  • One daily executive summary across all seven - punctuality, Bradford, checklists, training, RAs in one place.
  • Seven specialist AI managers on the data you already capture - no other UK suite ships this.
  • Candidate scoring on anonymised CVs - confidence-scored, hiring manager decides.
  • Absence insights framed as wellbeing-conversation starters, signed off by name.
  • Rotas drafted from real demand + availability + qualifications, manager publishes.
Capabilities at a glance

Everything AI should do - and nothing it shouldn't.

AI Health & Safety Manager

Briefings on overdue Risk Assessments, COSHH, accident patterns and unreported RIDDOR - with action buttons.

AI Attendance Manager

Absence patterns, Bradford trends advisory, lateness clustering, shift-coverage gaps - built on real Time Clock data.

AI Checklist Manager

Compliance trends, RAG distribution, zones and teams missing or late, risk hotspots.

AI Scheduling Manager

Open shifts, coverage gaps, employee availability mismatches, fairness alerts across the rota.

AI Training Manager

Overdue courses, completion rates by role, skills gaps, compliance expiries surfaced as a briefing.

AI HR Manager

Absence alerts, disciplinary and capability cases flagged for a manager to review - never auto-actioned.

AI Recruitment Manager

Active vacancies, pending candidates, offer stage, shortlist quality - briefed daily.

AI Schedule Builder

Draft rotas from real demand, availability, qualifications and fairness - with five configurable presets and a slider per constraint.

Risk Assessment AI suggestions

Draft control measures and alternative methods that the assessor rates and saves.

COSHH substance identification

GPT-4 Vision identifies a substance from a phone photo, searches for the SDS, drafts the hazards - assessor confirms.

Candidate shortlist scoring

PII-anonymised CV scored against role requirements; hiring manager makes the final call.

Bradford absence insights

Pseudonymised narrative patterns presented as wellbeing-conversation starters - never disciplinary triggers.

Deep dive

Every capability, organised by how you'll use them.

Skim the themes on the left, or scroll the full set. Each capability surfaces the specific frustration it removes and a real-world use case.

01

The seven AI Department Managers

A specialist AI manager for every operational area

Seven AI Department Managers, each scoped to a single area of the operation and briefed on the data that area already captures. Health & Safety briefs on overdue Risk Assessments, COSHH due-soon, accident patterns by weekday/time/zone, and unreported RIDDOR; Attendance briefs on absence patterns, Bradford trends (advisory) and lateness clustering; Checklists briefs on compliance trends, RAG distribution and risk hotspots; Scheduling briefs on open shifts and coverage gaps; Training briefs on overdue courses and skills gaps; HR briefs on absence alerts and capability cases (advisory only); Recruitment briefs on active vacancies, shortlist status and candidate quality. Every briefing has action buttons that link to the module; none of them auto-action.

Use caseA multi-site operator opens the morning dashboard, scans the seven manager cards in 60 seconds, and assigns three follow-ups - without leaving the briefing.

A daily executive summary across all seven - the cross-module brief no other suite ships

Each AI Department Manager generates a briefing on its own schedule (default 3× daily for Health & Safety / Attendance / Checklists, 2× for Scheduling, daily for the rest), and a single homepage executive summary synthesises across all seven into a one-glance morning brief. This cross-module unification is the part no other UK workforce platform ships - most vendors bolt an AI assistant onto one or two modules; nobody we've evaluated runs a specialist briefing for every operational area and a daily synthesis across all of them. The summary refreshes via a 5-minute background sweep job; manual refresh is rate-limited (six per user per hour, twenty per business per hour). Briefings live in-app only - there is no AI-generated email digest, by design.

Use caseA people-and-operations manager at a 14-site hospitality group opens AssistantManagerHQ at 08:00, sees the executive summary in 60 seconds, drills into Health & Safety because a COSHH is due-soon, and clears it before the team meeting - without opening four other tabs.

Freeing General Managers and Duty Managers from the daily chase

The five things a UK General Manager or Duty Manager chases every day - who clocked in late, whose Bradford Factor score moved, which checklist didn't get done on time, whose training certificate has expired, which risk assessment is due for annual review - are exactly the five things the AI Department Managers surface daily, with a one-click action button per item. The Attendance Manager flags punctuality and Bradford trends; the Checklist Manager flags missing and late completions by zone and team; the Training Manager flags expiring certificates by role; the Health & Safety Manager flags RA renewals and COSHH due-soon. The manager who used to spend the first 90 minutes of every shift chasing those items now spends 90 seconds reading the brief - and the rest of the shift running the operation. A General Manager handed back two hours a day pays for the platform six times over.

Use caseA duty manager at a hospitality venue opens the morning brief, sees four punctuality flags, one expired Food Safety Level 2 certificate and a checklist that didn't run yesterday - and has all four cleared before the kitchen brief at 11:00.
02

Documents AI drafts, humans sign off

Risk-assessment control measures, drafted and rated

On any open Risk Assessment the assessor can request AI control-measure suggestions for an activity, alternative methods for hazard elimination, or a combined AI review of the whole assessment. Each suggestion comes back as a structured advisory card - the assessor rates it helpful or not, edits as needed and saves to the RA. The ratings train nothing automatically; they're an internal quality signal. Suggestions never auto-save to the assessment, and the AI never marks an RA as approved.

Use caseA facilities-management assessor opens a new Working at Height RA, requests AI suggestions, accepts six of nine control measures verbatim, edits two, rejects one - and the RA is signed off in 25 minutes.

COSHH substance identification from a phone photo

The COSHH module accepts a phone photo of a chemical label or container. GPT-4 Vision identifies the substance, extracts the manufacturer and product name, and returns a structured identification with a confidence score; if confidence is ≥ 0.7, the platform automatically searches external SDS databases and drafts the hazard profile. The assessor confirms the identification (or rejects it and types the substance manually) before any COSHH assessment is created. The Vision call sees only the photo and a structured prompt - no employee data, no business data.

Use caseA kitchen manager photographs a degreaser bottle, the platform identifies it as a known caustic cleaner, pulls the SDS, drafts the hazard list - and the assessor signs off in five minutes.

Job descriptions, interview scorecards and training questions

In Recruitment, AI drafts job descriptions (text, requirements, benefits) and interview scorecards (criteria + sample questions) from role metadata - each draft is editable text the hiring manager reviews before posting or using. In Training, AI drafts drag-order and matching-pair training questions from a course outline (schema-constrained JSON). What AI doesn't do here: it doesn't draft offer letters, it doesn't generate whole courses end-to-end, and it doesn't draft formal performance or disciplinary documents. We say so rather than overclaim.

Use caseA care-group HR manager opens a new Senior Carer vacancy, AI drafts the JD and a six-criteria scorecard, the manager edits and posts in 15 minutes - instead of an hour.
03

Pattern detection — advisory, never disciplinary

Candidate shortlist scoring on anonymised CVs

Recruitment AI scores candidates against role requirements with a confidence rating and a short reasoning narrative. Before any CV reaches the model the platform strips names, email addresses, postcodes, dates of birth and National Insurance numbers; the prompt the model sees contains only experience, education, skills and structured role criteria. The output is stored on the candidate record as advisory JSON (`ai_match_score`, `ai_reasoning`) - the hiring manager makes the final shortlist decision. No auto-rejection. No auto-progression. Equality Act 2010 and GDPR Article 22 are both honoured by design.

Use caseA 200-site facilities-management group has 187 applicants for 12 night-cleaning roles; the AI ranks each anonymised CV in 90 seconds, the hiring manager reviews the top 30, shortlists 18, and the decision trail is on the candidate record.

Absence pattern insights - pseudonymised, advisory, signed off

The Bradford Factor module surfaces AI narrative insights on an employee's absence patterns - "post-rest-day absence rate 3.2× peer average", "three Mondays in the last six weeks, no other day-of-week clustering" - generated weekly by a background job. The employee's name is replaced with a token (e.g. `[EMPLOYEE]`) before the prompt is sent; the model only sees structured pattern data and the rolling-period context. Each insight is presented on screen as a dismissible card that a reviewer acknowledges by name and time; the prompt explicitly frames insights as "conversation starters for wellbeing checks" and AI insights never trigger a disciplinary action.

Use caseA general manager sees an insight that a colleague has clustered three short absences post-rest-day, opens a wellbeing check (the disciplinary route is blocked by code until that's recorded), and finds out it's a childcare clash - resolved with a small shift swap, not a formal route.
04

The AI Schedule Builder

The AI Schedule Builder - drafts, not autopilot

The AI Schedule Builder generates a draft rota from real-world inputs: employee availability windows, position requirements (shift-block demand), employee qualifications and training, fairness constraints (equal shift distribution), preference weighting (employee shift preferences 0-5), H&S coverage importance (weighted 0-5) and optional Bradford / absence-risk inputs. Five configurable presets - Balanced / Quality-Focused / Cost-Focused / Safety-First / Employee-Friendly - set slider weights across the constraints, and the manager picks (or fine-tunes) the preset before generating. Output is an `AIScheduleDraftVersion` with confidence scores; the manager edits, regenerates or approves before publishing.

Use caseA 50-site retail operator with 380 employees runs the Builder on Friday afternoon with the Safety-First preset; the draft is ready in 90 seconds with 92% confidence, the regional manager edits four shifts and publishes for Monday.

A constraint solver under the AI - not just a language model

The Builder isn't pure language-model output. A local constraint solver runs the hard constraints (availability, qualifications, position capacity) and the AI layer is used for the fuzzy optimisation (preference weighting, fairness across the team, demand-curve smoothing). When the AI layer is unavailable, the solver still produces a deterministic draft; when both are available, the AI layer improves the draft against the active preset.

Use caseA care provider uses the solver to enforce "only Level-3-qualified staff on night shift" as a hard rule, then the AI layer optimises fairness - a combination no spreadsheet macro could match.
05

Safety and privacy guards

GDPR Article 22 baked in - no autonomous decisions

Article 22 of the UK GDPR restricts decisions made solely by automated processing where they have legal or significant effects on an individual. AssistantManagerHQ AI is built so that no code path lets any model decision affect an individual alone - every candidate score, absence insight, control suggestion, scorecard, schedule draft and HR briefing is presented to a named human who acknowledges (or rejects) the output before it has any downstream effect. The audit trail names the human and the timestamp. By design, you have no Article 22 exposure from using our AI.

Use caseA hospitality-group compliance lead can answer the ICO question "is any individual-affecting decision made solely by AI?" with "no, and here is the decision audit."

PII anonymisation before any model call

In the two places where AI processes personal data - Recruitment candidate scoring and Bradford Factor absence insights - names, emails, postcodes, dates of birth and National Insurance numbers are removed before the prompt is built. Recruitment scoring sees structured CV content (experience, education, skills, role criteria) with the candidate represented by a candidate ID; Bradford insights see structured pattern data with the employee represented by a `[EMPLOYEE]` token. The prompts are auditable; the anonymisation is enforced in code, not in policy.

Use caseA care-group DPO can review the recruitment AI prompt template and the absence-insight prompt template and confirm that no personal data leaves the platform.
06

Models and the data they see

Models and the data they see

AI calls go to two providers. OpenAI GPT-4 family (currently `gpt-4.1`) is used for the seven manager briefings, the executive summary, RA control-measure suggestions, COSHH hazard drafting, Recruitment job descriptions and scorecards, training question generation, Bradford absence insights, candidate scoring and the AI Schedule Builder's optimisation layer. OpenAI GPT-4 Vision is used for COSHH substance identification from a photo. Azure Document Intelligence (Form Recognizer) is used for OCR - accident-form handwriting extraction and CV text extraction. Other models (Claude, Gemini, local LLMs) are not used. Switching providers is a configuration change; switching is roadmap, not shipped.

Use caseA regulated buyer (care, education, healthcare) can put the model list in the IG questionnaire and the DPO can sign off.
FAQ

Questions buyers ask before they switch.

Answers from our UK-based onboarding team. Need something not covered? Book a walkthrough →

Two specific things. One - seven specialist AI managers in a single suite, with a cross-module executive summary. Most platforms ship one or two AI assistants bolted onto one or two modules (BambooHR's chatbot, Lattice's review summariser, Sage's payroll prompts). *We haven't found another UK workforce platform that runs a specialist briefing for every operational area - Health & Safety, Attendance, Checklists, Scheduling, Training, HR, Recruitment - and* synthesises across all seven into a single daily brief. That cross-module unification is the part of the design we believe is unique. Two - advisory by code, not by policy.** Most AI HR tools are built to autonomously act (auto-reject candidates, auto-flag disciplinaries, auto-email staff). Ours is built so no code path lets AI affect an individual without a named human signing off. GDPR Article 22 is a design constraint, not a footnote.

One AI suite. Seven specialist managers. Two hours back per shift for your GMs - start this afternoon.

Seven AI managers + cross-module briefFrees GMs from the daily chaseGDPR Article 22 by designNamed human signs off every decision