Half day audit
Half a day£850
FitSmall UK team, one specific automation or AI feature you want a second opinion on before spending real money.
OutputWritten review, recommended stack, honest go or no go with reasoning.
AI Automation Engineer, UK
Hire the engineer, not the pitch deck.
I am Bishal Paul, an independent AI automation engineer based in the UK. I take a small number of scoped engagements per quarter and ship each one into real production. You speak to me from discovery through to the runbook.
10 yrs
shipping software
4 yrs
AI automation focus
6+ sectors
in production
4 products
live under own name
Hi, my name is Bishal Paul. I am an AI automation engineer working with UK teams that need real production systems, not slide deck demos. I run Erudience (see erudience.com), the AI automation agency I founded, and I am Head of AI at Absolute Intelligence UK. Under my own name I also ship four live products: push8, Extinde, Automation Workflows and ExcelErrorFinder. Every one of them runs in production, which is where I do my strongest work.
The UK AI automation market is loud. Agencies resell tools, junior implementers hand off demos, and nobody stays around long enough to see what breaks on a Tuesday morning. My work is the opposite: I sit inside the failure modes a real production system has to survive. Late FTP feeds. Retry loops. LLM calls that time out under load. Phone calls that hit voicemail at exactly the wrong moment.
Sectors already in production
Where these systems already live for UK teams.
- § 01Telecom
Outbound voice collections with in call card capture.
- § 02Retail and e commerce
Voice ordering into Shopify, seven department inbound routing.
- § 03Energy
Multilingual subsidy qualification on a single call.
- § 04Local services
Weekend and overflow coverage priced per minute, not per seat.
- § 05Financial services and M&A
Deal sourcing and LOI drafting with Slack approval on every outbound.
- § 06Agency ops
Multi brand social pipelines across LinkedIn, Facebook, Instagram.
UK market context
Three numbers worth having in the room.
£21bn+
estimated UK AI sector revenue with 3,700+ companies employing over 60,000 people.
Source · DSIT UK AI Sector Study 2024
72%
of organisations globally report using AI in at least one function, up from 55% the prior year.
Source · McKinsey State of AI 2024
>30%
of generative AI projects expected to be abandoned after proof of concept by end of 2025.
Source · Gartner GenAI Forecast, July 2024
Independent engineer vs agency
How the two shapes actually differ in practice.
Area
Independent engineer
Agency
Who you speak to on day one
The engineer building your system
A sales lead, then an account manager
Who does the build
The engineer, plus specialists on demand
Junior implementers behind the pitch team
Pricing shape
Fixed scope, fixed price, weeks not months
Time and materials on a rolling retainer
What happens after launch
Optional retained operator, cancellable monthly
Same retainer, harder to reduce
Willingness to say no
Turns down builds that will not survive prod
Sells the biggest scope the client will sign
How I engage
Five stages, in this order.
No open ended time and materials. Scope, timeline and price are agreed before the build starts. The plan below is the same for every engagement.
- 1
Discovery call
30 minutes on the actual problem, current state, and what a good outcome looks like. Free.
- 2
Paid scoping
1 to 2 weeks. Produces a clear architecture, a fixed build price, and a timeline you can commit to.
- 3
Build
End to end, with weekly demos into a shared environment. You see the workflow taking shape, not a reveal at the end.
- 4
Launch and handover
Into your infrastructure. Runbook, monitoring, and a session with your operator.
- 5
Retained operator
Optional. A defined block of hours per month for evolution, incidents and second opinions. Cancellable monthly.
Tools in use
The stack I default to, grouped by job.
Automation runtime
- n8n cloud
- n8n self hosted
- Make
- Zapier
- Pipedream
- Windmill
Voice + telephony
- ElevenLabs
- Vapi
- Retell
- Twilio
- Vonage
- OpenAI Realtime
- Deepgram
LLM + reasoning
- Claude 3.5 / 4
- GPT-4o / GPT-5
- Gemini 2
- Groq
- Structured outputs
- Function calling
Data + storage
- Supabase
- Postgres
- Redis
- Pinecone
- Qdrant
- pgvector
Ops + delivery
- Docker Compose
- DigitalOcean
- Hetzner
- Cloudflare
- GitHub Actions
- Sentry
- Grafana
Observability + eval
- Langfuse
- LangSmith
- Helicone
- Prompt versioning
- Golden sets
- Cost telemetry
Engagement models
Engagement models and pricing for UK teams.
The first 30 minute discovery call is free. Beyond that, every engagement is scoped, priced, and dated before the build starts. Numbers below are indicative floor prices in GBP, exclusive of VAT and third party costs (LLM tokens, telephony minutes, infrastructure). No hourly time and materials.
Architecture sprint
1 to 2 weeksFrom £1,800
FitYou have a problem and need an honest architecture, a fixed build price, and a timeline before committing.
OutputArchitecture doc, integration map, risk register, fixed build quote.
Fixed scope build
3 to 8 weeksFrom £8,500
FitYou have a defined workflow to ship. Usually one voice or automation system, in production, on your infrastructure.
OutputLive system, runbook, monitoring dashboards, operator handover session.
Retained operator
Rolling monthlyFrom £2,800 / month
FitThe system is live and you want a defined block of engineering hours for evolution, incidents, and second opinions.
OutputWeekly working session, on call for defined severity, quarterly review of cost and reliability.
Advisory retainer
Rolling monthlyFrom £950 / month
FitYou already have an in house team, you want a senior AI engineer as a sounding board, not a builder.
OutputAsync architecture reviews, procurement help on tools, hiring rubric for AI roles.
Signals it is time to hire
Six signs the problem is now bigger than a Zapier fix.
If two or more of these are already true, the ROI on a scoped engagement is usually clear inside the discovery call.
- 01
Support inbox, outbound calls, or manual data work is scaling headcount linearly with revenue.
- 02
You already tried Zapier or Make and hit the ceiling on branching, retries, or state.
- 03
An LLM feature works in a notebook but nobody on the team is comfortable putting it in front of paying users.
- 04
A contractor built a proof of concept months ago and it still is not in production.
- 05
You need after hours or weekend coverage on phone or chat without adding shift staff.
- 06
Finance is asking why the AI bill is unpredictable and nobody can answer.
Integrations already in production
The systems most UK teams need to plug into.
Not a marketing checklist. Every tool below is one I have already wired into a live client workflow, usually more than once.
CRM + sales
- HubSpot
- Salesforce
- Pipedrive
- Attio
- Close
Support + ticketing
- Zendesk
- Intercom
- Freshdesk
- Front
- Help Scout
Commerce
- Shopify
- WooCommerce
- Stripe
- Recharge
- Klaviyo
Comms + ops
- Slack
- Microsoft Teams
- Gmail
- Outlook 365
- Notion
- Airtable
Voice + telephony
- Twilio
- Vonage
- 8x8
- Aircall
- GoTo Connect
Docs + finance
- DocuSign
- Xero
- QuickBooks
- Google Drive
- SharePoint
When I am not the right hire
Five cases where I say no.
Saying no early is part of the job. If any two of these describe your situation, we should talk anyway, but the honest recommendation will not be an engagement.
- The underlying process is broken or undocumented. Automation will just make failure faster.
- Volume is tiny. A spreadsheet, a well written SOP or a single scheduled script is cheaper.
- The real problem is org design, ownership or incentives. No system fixes that.
- You want a slide deck vendor, a monthly retainer with no defined output, or an unlimited scope experiment.
- You need a full time employee, not an engagement. I will say so and point you at the right hire.
Recent production work
Three systems recently shipped into live use.
Case study
Outbound voice collections with card capture
Telecom outbound with DTMF payment handoff and structured call outcomes.
Read case study
Case study
Multilingual subsidy qualification agent
Energy client, cross lingual eligibility flow validated on live calls.
Read case study
Case study
M&A deal sourcing and LOI drafting
Sourcing, qualification, LOI draft, Slack approval on every outbound.
Read case study
Answered before the call
Nine questions I get every week.
What does an AI automation engineer actually do?+
Design, build and operate systems that use AI models plus workflow tooling to remove repetitive work. In my case: self hosted n8n on DigitalOcean, Supabase for state, Claude and OpenAI for reasoning, ElevenLabs and Twilio when the workflow involves phone calls. Most of the value sits in the integration and operational layer, not the model.
Are you UK based, and do you work outside the UK?+
Yes to both. Erudience is a UK entity. Most client work is with UK teams, a meaningful share is US and EU. Remote is default, on site is available for UK clients when it actually helps the build.
How is this different from hiring an agency?+
Agencies resell tools through junior implementers. Every engagement I take is delivered by me end to end, or by a small trusted network I bring in for specific specialisms. You always speak to the person building the system.
What is a typical engagement size?+
Fixed scope discovery, a defined build project measured in weeks not months, or a retained operator role on a system already in production. No open ended time and materials.
How do you decide what to automate and what to leave alone?+
Discovery starts with the process, not the tools. Repetitive, rule based, high volume, clear success signal, that is a strong candidate. Judgement calls, ambiguous inputs or a hidden broken process, the answer is often to fix the process first. Saying no is part of the job.
When is hiring an AI automation engineer the wrong move?+
If the underlying process is broken or undocumented, automation makes it faster to fail. Fix the process first. If the volume is genuinely tiny, a spreadsheet or a well written SOP beats any system. And if the real problem is org design, not tooling, an engineer will not solve it. I will say so at scoping.
Do you sign NDAs, DPAs and work with regulated sectors?+
Yes. Mutual NDA at first call if requested, DPA before any client data moves. I ship into financial services, energy and professional services, all with structured audit trails, data residency respected, and recordings stored under the client's policy.
What is a realistic timeline from first call to live system?+
Free discovery in the same week. Paid scoping in one to two weeks after that. Build in three to eight weeks depending on scope. So four to twelve weeks end to end for a single scoped system, and the first weekly demo lands seven to ten days into the build.
Can you take over a system another team built?+
Yes, and it is one of my most common opening engagements. Health check first: what runs, what does not, what is undocumented, where the risk sits. Then a written recovery plan, then a fixed scope stabilisation, then optional retained operator. I will not silently rewrite things without agreement.
Bishal Paul
Founder, Erudience
Head of AI, Absolute Intelligence UK
Available for new work
UK based, working across timezones
Reply within 24 hours
Start a conversation
Let's scope your automation build.
A free 30 minute discovery call. You leave with a concrete direction on scope, sequencing and cost, even if we do not end up working together.
- Direct line to me, not an account manager or a sales team.
- No slide deck. We look at your current workflow and where it hurts.
- Written summary in your inbox within 24 hours of the call.