Services
Seven engagements, in three families. All of them written down before you pay.
Each has numeric scope caps, a fixed price agreed in writing before work starts, and a timeline that begins when access is complete rather than when the contract is signed. And every engagement that teaches a person anything is delivered to exactly one person at a time.
Or answer two questions, and get routed the way a fit call would route you.
The same logic the first two minutes of the call runs — including the honest exit for the cases this practice should not take.
What are you trying to change?
Every path, written out
What are you trying to change?
- One person’s capability. Mine, or one named colleague’s — how one professional actually uses AI in their own work. → Solo Track. One person, their own intake, their own roadmap, six one-to-one sessions on their own live work.
- A whole team’s capability. Several named people — each needs to be genuinely better at this, not present at a session about it. → Solo Rollout. One programme per named person — every individual gets their own intake, roadmap and private sessions, and the sponsor sees aggregate progress.
- A process that eats hours. Something your team does by hand every week that a system should be doing. → one more question, below.
- An operation nobody can fully describe. The process document and the reality parted ways years ago — someone has to watch the work to map it. → Forward Deployed Engineer. An engineer placed inside the operation for a month or more, mapping the work as it actually runs before building the AI layer for it.
- A business that has not launched yet. Pre-revenue, pre-process — exploring what AI could do for a company that is still being built. → Honestly — not yet. A build needs a process that already runs and already costs counted hours, and there is no way to scope one against a business that does not exist yet. Spend the money on customers first. When a process starts eating real hours every week, the fit will be obvious — and one professional sharpening their own capability is always a legitimate engagement, with or without a company behind them.
How well is that process understood?
- We can write down every step. The inputs, the rules, the systems it touches — known today. → Workflow Automation Sprint. A known process with known systems is a fixed-scope build: live in ten working days from Access-Complete.
- It is answering the same questions. Support, sales or internal queries answered again and again from documents and past tickets. → AI Agent Build. Repeated questions over a corpus is retrieval work — an agent with an evaluation suite, not a workflow.
- Several processes, tangled together. More than one workflow, shared systems, and no clean seam between them. → Ops Automation Package. A bundled scope with the seams mapped first, so the automations land in an order that compounds.
- We know it hurts, not where the return is. There is obvious waste, but ranking what to automate first is the actual question. → AI Opportunity Diagnostic. Five days, a written process map, and three to five ranked candidates with the assumptions stated — credited in full against any build within thirty days.
One person at a time, always — because nobody changes how they work by watching somebody else be taught.
How that works →For individuals, and for every person in a team
Solo AI adoption
Enablement delivered privately, on the buyer’s own live work. Bought by a professional for themselves, or by a sponsor for a whole function — in which case it becomes one programme per person rather than one session for everybody.
One person at a time. Never a group.
The dedicated page →Solo Track
One professional, one written roadmap built around their own work, and six one-to-one sessions on the tasks they actually do.
A programme for exactly one person. It opens with an intake into how your week genuinely runs, produces a written map of where AI belongs in it and where it does not, then works through that map with you on your own live files. Nobody else is in the session, which is the point — no question is too basic to ask.
What you get
- A 60-minute intake into how your week actually runs, task by task
- A written personal AI map: which recurring tasks are worth moving, which are not, and in what order
- Six one-to-one working sessions on your own live files, never on exercise data
- A personal SOP pack — the prompts, the checks, and the rules for when not to use a model at all
- Recordings of every session, kept for you and shared with nobody else
- A closing review against the map, so you can see what changed and what to practise next
- A 30-day written follow-up channel after the final session
Best fit
You are an accountant, a recruiter, a lawyer or a support lead who wants to be genuinely good at this, and would rather work on your own files than sit through a generic curriculum.
What is fixed
- Timeline
- 4–6 weeks, scheduled around your working week
- Payment
- 50% to start / 50% at the final session
- Price
- Quoted after a 20-minute fit call, fixed in writing before work starts.
- Scope caps
- 1 person; the programme is not transferable to a colleague
- 1 intake (60 min) and 6 one-to-one working sessions (60 min)
- Up to 8 of your own recurring tasks rebuilt
- 30 days of written follow-up after the final session
Twenty minutes. No advice given, no pitch.
How it runs
4–6 weeks, scheduled around your working week
Your intake
Day 0 · 60 minutes, nobody else on the call
A normal week of yours, task by task: what recurs, which tool it happens in, and the parts you quietly avoid.
Your written map
Within 3 business days of the intake
Which of your recurring tasks are worth moving, which are not worth touching, and the order to take them in. You read it first and can strike anything on it.
Six one-to-one sessions
Weeks 1–6 · your own live files, never exercise data
Up to eight of your own recurring tasks rebuilt with your hands on the keyboard. No other learner is present, at any point.
Review, SOP pack, follow-up
Then 30 days of written follow-up
A closing review against your own map, a personal SOP pack in the vocabulary of your job, and a channel for the questions that only appear in week two.
Solo Rollout
Your team raised one person at a time — a separate intake, a separate roadmap and separate sessions for every individual.
An enterprise engagement that never becomes a class. The sponsor states what the organisation needs to change; then every named individual gets a private intake, a roadmap built on how that person actually works, and their own one-to-one sessions. Twelve people means twelve programmes running in parallel, and the sponsor sees progress in aggregate only.
What you get
- A sponsor alignment call that records what the organisation expects this to change
- A private 45-minute intake with every named individual, run one at a time
- A written roadmap per person, built on how that person’s own week actually runs
- Four one-to-one sessions per person on their own live work, scheduled to that person’s calendar
- A personal SOP pack for each individual, written in the vocabulary of their own job
- Aggregate-only reporting to the sponsor: adoption and themes, never who struggled with what
- A consolidated automation-candidate register — the by-product of asking twelve people how they really work
Best fit
A sponsor who needs a whole function to become competent, and has watched a single shared session leave the quietest half of the team exactly where they started.
What is fixed
- Timeline
- 6–11 weeks by headcount, plus 30 days of follow-up
- Payment
- One purchase order for the whole rollout · 50% to start / 50% at completion
- Price
- Quoted after a 20-minute fit call, fixed in writing before work starts.
- Scope caps
- 1 sponsor alignment call (60 min) before any individual is contacted
- Per person: 1 intake (45 min) and 4 one-to-one sessions (60 min)
- Up to 12 individual programmes in delivery at any one time; larger rollouts run in sequenced waves
- 2 aggregate progress reports to the sponsor; no individual transcripts
Twenty minutes. No advice given, no pitch.
How it runs
6–11 weeks by headcount, plus 30 days of follow-up
Sponsor alignment
Week 1 · 60 minutes, before anyone is contacted
The sponsor states what has to be true in ninety days, and what the organisation will not fund.
One private intake each
Weeks 1–3 · 45 minutes per named person
Run one at a time. What is said in an intake is never repeated to a manager or a sponsor.
One roadmap per person
Week 3 · different starting points, one standard
Written against how that individual actually works. Each person reads their own before anyone else does.
Four sessions per person
Weeks 4–10 · up to 12 programmes in parallel
Twelve people is twelve separate deliveries, each on that person’s own live work and their own calendar.
Aggregate readout
Week 11, then 30 days of follow-up
Coverage, themes and blockers to the sponsor. No transcripts, no ranking, and no note of who found what hard.
Why none of these is a group format
What group training does well, and what it cannot do
Group training is not a fraud, and pretending otherwise would be dishonest: it is genuinely cheap per head, genuinely fast to schedule, and the right instrument for announcing a policy or a new tool. It is simply not an instrument for changing how one named person works on Monday morning.
Cost per head
A group class: Genuinely low, and it falls further with every extra person, because one delivery covers all of them.
A SoloZ solo track: Quoted per person, because it is delivered per person. That is the honest trade for a programme built around one named week.
Who the content is built for
A group class: The median of the people booked — defensible, consistent, and precisely nobody's actual job.
A SoloZ solo track: One named person's tasks, tools and habits, collected in an intake before a single slide or example is prepared.
The least confident person
A group class: Stays silent. Asking the basic question in front of a manager and peers costs more than not knowing the answer.
A SoloZ solo track: Asks it in the first ten minutes, because there is no one to be measured against and no impression being formed.
The week afterwards
A group class: A shared handout and an open channel. Whoever was already confident applies it; the rest return to old habits.
A SoloZ solo track: Follow-up sessions against that person's real backlog, with the next task to convert already named and agreed.
One day, then nothing. Or one step at a time, on your own work.
The vertical axis is working proficiency — what someone can actually do on a Tuesday, not what they can recall from a slide.
- Solo track
- One shared day
Read this chart as a table
| Week | Solo track | One shared day |
|---|---|---|
| Start | Curious | Curious |
| Week 1 | Occasional | Occasional |
| Week 2 | Occasional | Occasional |
| Week 3 | Regular use | Occasional |
| Week 4 | Regular use | Occasional |
| Week 6 | Systematic | Occasional |
| Week 8 | Systematic | Occasional |
| Week 10 | Systematic | Curious |
| Week 12 | Systematic | Curious |
Every engagement on this page that teaches a person anything is delivered to exactly one person at a time. That is the reason there is no fourth family here, and the reason a rollout for twelve is quoted as twelve programmes rather than as one delivery.
The solo model in fullFor the business
Build & automate
A paid diagnostic that ranks what to automate, then fixed-scope builds: workflow automation, retrieval-backed agents with an evaluation suite, or a full operations package. Version-controlled, tested, and handed over with a runbook.
Production software, in infrastructure you own.
The dedicated page →AI Opportunity Diagnostic
A ranked, costed map of what to automate first — before you commit to building anything.
Five business days, one department, a written diagnostic you could hand to any engineer. It exists to replace the free consulting call that wastes both our time, and its fee is credited in full against a build commissioned within thirty days.
What you get
- An 8–12 page written diagnostic, not a slide deck
- Process map of one department, as it actually runs today
- 3–5 automation candidates, ranked by return
- Hours-saved estimate per candidate, with the assumptions stated so you can challenge them
- Recommended stack and a realistic effort estimate
- 90-day sequencing plan — what to do first, second, and not yet
- 45-minute walkthrough call to defend the reasoning
Best fit
You know something is wasting your team’s time but you cannot yet write a scope for it. Start here.
What is fixed
- Timeline
- 5 business days
- Payment
- Paid in full upfront · credited against any build commissioned within 30 days
- Price
- Quoted after a 20-minute fit call, fixed in writing before work starts.
- Scope caps
- 1 department
- 3–5 ranked automation candidates
- 1 walkthrough call (45 min)
Twenty minutes. No advice given, no pitch.
How it runs
5 business days
Process capture
Day 1 · one department
How the work is done today, recorded from the people who do it rather than from the org chart.
Candidates ranked
Days 2–3 · 3–5 automation candidates
Each with an hours-saved estimate and the assumptions stated in writing, so you can challenge them.
Stack and sequencing
Day 4 · a 90-day plan
What to do first, what to do second, and what is not worth touching yet — with the reasoning attached.
Document, then a walkthrough
Day 5 · 8–12 pages, then a 45-minute call
A written diagnostic rather than a deck. You own it either way, including if you take it to another engineer.
Workflow Automation Sprint
One process your team does by hand every day, running without them in ten working days.
A single business process, built as production software in infrastructure you own, with error handling and a runbook. Fixed scope, fixed price, written before you pay.
What you get
- A production workflow running in your own infrastructure
- Error handling, retries, and a dead-letter path for everything that can fail
- A written runbook your team can operate without me
- A recorded walkthrough of how it works and how to change it
- 14-day post-launch bug window
- Everything in version control, in your repository
Best fit
You can name the process and count the hours it costs you every week. Two people, three hours a day, copying between systems.
What is fixed
- Timeline
- 10 working days from Access-Complete
- Payment
- Milestone-based · agreed in the statement of work before work begins
- Price
- Quoted after a 20-minute fit call, fixed in writing before work starts.
- Scope caps
- 1 business process
- Up to 3 system integrations
- 2 revision rounds included
Twenty minutes. No advice given, no pitch.
How it runs
10 working days from Access-Complete
Access-Complete
Day 0 · the ten working days start here, not at signature
The one process is named and its up-to-three system integrations are listed, with credentials, a test environment and a signed definition of done in hand.
Every branch, including month-end
Days 1–2 · you sign the map before anything is built
The process written down as it is really done, exceptions included — the ones that surface only at quarter end are where an automation quietly fails.
Built, then broken on purpose
Days 3–8 · committed to your repository as it is written
It goes together in your own infrastructure, then runs against bad data, dead integrations and timeouts until the retries and the dead-letter path hold.
Runbook, walkthrough, handover
Days 9–10 · then a 14-day bug window
A written runbook and a recorded walkthrough, so your team operates and changes it without me. Two revision rounds are included against the signed map.
AI Agent Build
An assistant that answers from your documents, knows when to escalate, and is tested before you trust it.
A grounded, retrieval-backed agent on one channel — and, unusually at this level of the market, it ships with an evaluation suite and a documented pass threshold, so "is it working?" has a numeric answer rather than an opinion.
What you get
- Retrieval pipeline with the chunking strategy documented, not guessed
- Grounded answers with citations back to your source documents
- An escalation-to-human path for everything the agent should not answer
- Evaluation suite: 40 golden question/answer pairs, a documented pass threshold, and a regression run at handover
- Admin console so your team can update the corpus without calling me
- Cost-per-conversation projection so the running cost is not a surprise
Best fit
Your support inbox gets the same two hundred questions a week, or your team keeps searching the same document set to answer customers.
What is fixed
- Timeline
- 10–15 working days from Access-Complete
- Payment
- 50% to start / 50% at UAT
- Price
- Quoted after a 20-minute fit call, fixed in writing before work starts.
- Scope caps
- 1 agent, 1 channel (web, WhatsApp, Slack or email)
- Up to 2 integrations
- Corpus up to 1,000 documents or 5,000 pages
Twenty minutes. No advice given, no pitch.
How it runs
10–15 working days from Access-Complete
Access-Complete
Day 0 · the delivery clock starts here
The corpus handed over — up to 1,000 documents or 5,000 pages — plus a named owner, a test environment and a signed definition of done.
Corpus and retrieval
Days 1–3 · chunking documented, not guessed
Your documents indexed, with the strategy written down so somebody else can change it later.
Grounded answers
Days 4–9 · citations back to your sources
Answers tied to the document they came from, plus the escalation path for everything the agent should refuse.
Evaluation suite
Days 10–13 · 40 golden pairs, documented threshold
Below the documented threshold is not shipped. “Is it working?” gets a number instead of an opinion.
UAT and handover
Days 14–15 · admin console, running-cost projection
Your team updates the corpus without calling me, and knows what a conversation costs to run before launch.
Ops Automation Package
The full operational layer — several workflows, an agent, and a map of how they fit together.
For teams past the point of automating one thing. Three workflows, one agent, a documented integration map, and a handover that leaves your team able to run all of it — delivered, like everything else here, one person at a time.
What you get
- Three production workflows, each with its own runbook
- One grounded agent with an evaluation suite
- A documented integration map of how your systems now talk to each other
- A recorded one-to-one handover for each person who has to operate it
- 14-day post-launch bug window across the whole package
Best fit
Operations is the bottleneck across more than one function, and you want it addressed as a system rather than one ticket at a time.
What is fixed
- Timeline
- 4–6 weeks
- Payment
- Three milestones · agreed in the statement of work before work begins
- Price
- Quoted after a 20-minute fit call, fixed in writing before work starts.
- Scope caps
- 3 workflows
- 1 agent
- 1 handover session per named owner
Twenty minutes. No advice given, no pitch.
How it runs
4–6 weeks
Integration map
Week 1 · agreed before anything is built
How your systems talk to each other today, and exactly where the three workflows will sit between them.
Workflows one to three
Weeks 2–3 · each with its own runbook
Built and tested one at a time, so a delay on one does not hold the other two behind it.
The agent and its evals
Weeks 4–5 · a documented pass threshold
Grounded on your own documents, with a regression run before it goes anywhere near a customer.
One handover per named owner
Week 6 · recorded, one to one
Pitched at what that person actually has to do with it. A 14-day bug window covers the whole package.
For operations that cannot be scoped from outside
Forward deployed
An engineer placed in your office or entirely inside your working hours. The first week is spent watching how the job is really done; what follows is built into your own infrastructure and handed over one person at a time.
An engineer inside the work, for a month or more.
The dedicated page →Forward Deployed Engineer
An engineer placed inside your operation for a month or more, building the AI layer where the work actually happens.
Some work cannot be scoped from the outside. A forward deployed engineer is placed with you — in your office or inside your working hours — for a month or more, watches the work as it is done rather than as it is described, and builds the AI layer into your own infrastructure while sitting beside the people who will have to run it.
What you get
- Week one spent watching the work as it is actually done, ending in a written map of it
- An AI layer built in your own infrastructure and your own repositories, in version control
- An evaluation suite on everything that touches a language model, with a documented pass threshold
- Retries, timeouts, dead-letter paths and failure alerting into a channel your team already reads
- A runbook and a recorded walkthrough per component, written for whoever inherits it
- A weekly written note to your sponsor: what shipped, what is blocked, what changed in the plan
- One-to-one handover to each person who has to operate what was built
Best fit
The thing you need automated cannot be written as a scope document, because nobody outside the operation understands how it truly runs. Have someone come and watch it for a month.
What is fixed
- Timeline
- 1 month minimum, extended in monthly blocks
- Payment
- Monthly, invoiced in advance · 30 days’ written notice on either side
- Price
- Quoted after a 20-minute fit call, fixed in writing before work starts.
- Scope caps
- 1 engineer placed with 1 client, minimum 1 month, extended in monthly blocks
- 4 days a week inside your working hours; 1 day held for build and review
- Scope frozen at the end of week 1, then re-agreed at each monthly renewal
- On-site within Bengaluru; elsewhere remote by default or quoted with travel
Twenty minutes. No advice given, no pitch.
How it runs
1 month minimum, extended in monthly blocks
- 1
Week 1
Watching the work
Four days a week inside your working hours, next to the people doing the job — not reading a process document written about it.
Exit gate · A written map of how the work is actually done, and the scope frozen against that map.
- 2
Week 2
The first component
The costliest step on the map, built into your own infrastructure and committed to your own repository as it is written.
Exit gate · It runs against real work, in version control, with retries and a dead-letter path.
- 3
Week 3
Failure paths and evals
The bad day, rehearsed: timeouts, rate limits, bad data — plus an evaluation suite on anything that touches a language model.
Exit gate · A documented pass threshold met, and failure alerting into a channel your team already reads.
- 4
Week 4
Handover, then renew or stop
A runbook and a recorded walkthrough per component, handed to whoever inherits it — one person at a time, as everywhere else here.
Exit gate · Scope re-agreed in writing before a further month starts, or the placement ends with the repository and the runbooks already yours.
The part nobody prices in
Confidence is a delivery constraint, not a soft concern
The person who most needs the training is almost always the person least willing to be seen needing it. Put them in a session with eight colleagues and a manager present, and the arithmetic is immediate: asking the basic question costs more socially than not knowing the answer costs professionally. So they stay quiet, nod in the right places, and leave with exactly the habits they arrived with. The delivery is recorded as complete. Nothing has changed.
This is not a matter of temperament, and it is not fixed by announcing that there are no stupid questions. It is fixed by removing the audience. One to one, there is no colleague to be ranked against and no manager forming a view, so the first question asked is usually the one that was actually blocking the work — how to open the thing, what the model is doing with the document, why yesterday’s attempt produced nonsense. Those questions are never asked in company, and they are the ones that matter.
Where that shows up in this catalogue
- Every adoption session
- Solo Track and Solo Rollout are one person and one engineer. There is no larger format to ask for, and none appears under deadline.
- Every build handover
- A separate recorded session for each named owner, pitched at what that individual has to do with it — never one briefing everybody sits through.
- Every forward deployed placement
- The system is handed to each named owner in their own recorded session, and the weekly note to your sponsor records what shipped, never who asked what.
Twelve is the Solo Rollout cap printed above: twelve programmes in delivery at once, one sponsor, one engineering standard, and no line joining any two learners to each other. Scale here means running the method more times, never putting more people in one delivery.
What keeps it fixed
Three clauses that do most of the work.
Everything above is only meaningful if the scope and the clock behave the way the page says they do. These are the mechanisms that make them.
Numeric scope caps
On the statement of work
Two revision rounds are included in every build. A third is a change order, quoted and agreed in writing before any work starts.
Change orders, priced to discourage
Agreed before anything happens
Anything outside the written scope caps is quoted before it happens, at the change-order rate stated in your statement of work. That rate sits deliberately above the blended project rate, so it is always cheaper for you to scope a new piece of work properly than to extend an existing one informally.
The clock starts at Access-Complete
Not at signature
Every build timeline is quoted in working days from the point at which credentials, a named owner on your side, a test environment, real sample data, an available subject-matter expert and a signed definition of done are all in hand. Most slipped deadlines in this industry are access delays counted as build time.
And on the adoption side
Adoption engagements are quoted per person, because that is how they are delivered. A rollout for twelve people is twelve programmes, and the quote says so rather than hiding the headcount inside a day rate.
What it costs
Every engagement is quoted individually — and quickly.
There is no price list here because there is no honest one. What a workflow, an agent or a rollout costs depends on how many systems it touches, how clean the data is, how much of the process is written down, and — for adoption work — how many people are actually in it. A number quoted before knowing those is either padded to cover the unknown or discovered to be wrong halfway through.
What you get instead is a fixed price, in writing, before you commit to anything — and fast enough that it is never the reason a project stalls.
From first contact to a number
- 01
A reply within 24 hours
From the person who would actually do the work, not an automated sequence.
- 02
A 20-minute fit call
Qualification only. Which process or which people, how many hours a week, and who signs off. No advice given and no pitch.
- 03
A written proposal within 24 hours of that call
A fixed price, with the scope caps, timeline, deliverables and payment schedule already in it.
- 04
The price then does not move
Two revision rounds are included on builds. Anything outside the written caps is quoted and agreed before it happens, never billed as a surprise.
How it works
Four steps, and you can stop after any of them.
There is no long discovery phase and no retainer you have to sign before anything happens. Each step produces something you keep.
Fit call
20 minutes · free
A qualification call, not a consulting session. Which process, how many hours a week, who signs off, what you have already tried. If it is not a fit, I say so on the call rather than sending a proposal.
Diagnostic
5 business days · paid
A written diagnostic that ranks what to automate and what it will return. Its fee is credited in full against a build. You own it either way — including if you take it to another engineer.
Build
10 working days from Access-Complete
Fixed scope and fixed price agreed in writing before anything starts. Built in your infrastructure, in version control, tested against an evaluation suite, and handed over with a runbook.
Care
Monthly · optional
Monitoring, tuning, regression runs, and model migrations when a vendor deprecates something underneath you. Optional, and easy to cancel — the system is yours regardless.
Inside a build
What actually happens between the deposit and the handover
Every build follows the same path, and each stage has a gate that has to be met before the next one starts. It is unglamorous and it is the reason the ten days stays honest.
Access-Complete
Credentials, a named owner on your side, a test environment, real sample data and a signed definition of done. The delivery clock starts here, not at signature.
Process map frozen
Edge cases enumerated and acceptance criteria agreed in writing. You sign the map before a line of it gets built.
Modular build
Small components rather than one monolithic flow, configuration externalised, everything committed to your repository as it is written.
Failure injection
Bad data, dead integrations, timeouts and rate limits, run deliberately against the happy path until the retries and dead-letter paths hold.
Evaluation run
Forty golden pairs against a documented pass threshold on anything touching a language model. Below threshold is not shipped.
One-to-one handover
A separate recorded session for each named owner, pitched at what that person has to do with it, plus the runbook and the 14-day bug window.
What production-grade means here
Most AI automation is a prototype
with a confident tone.
The demo always works. The question is what happens on the bad day — when an API times out, a model is deprecated, or someone asks the assistant something it should refuse. That is the part I get paid for.
These are not upsells or a premium tier. They are the standard every engagement is delivered to, and they are the reason a quote is what it is.
You own the infrastructure
Everything runs in your cloud, your automation instance, your API keys, your vector database. I get delegated admin access that you can revoke in one click. You are never renting your own operations back from me.
Everything is in version control
Workflows exported to your Git repository. Prompts as versioned files, not pasted into a web form and forgotten. You can read the history, diff a change, and roll it back.
Every AI deliverable ships with an evaluation suite
Forty golden question/answer pairs, a documented pass threshold, and a regression run before handover and before any change. When someone claims the assistant "got worse", there is a number to check instead of an argument.
Failure paths are built, not assumed
Retries, timeouts, dead-letter handling, and alerting into a channel your team already reads. The interesting part of automation is what happens on the bad day, not the demo.
Observability from day one
Execution logging, failure alerting, and a token-cost view — shipped with the build rather than sold back to you later as an upgrade.
Handover is one to one, like everything else
Each person who has to operate or change the system gets their own recorded session, pitched at what that individual actually needs to know. Nobody is asked to keep pace with anyone else in a shared briefing.
Where it lands
Which industries does this actually work for?
The ones that run on high-volume, rule-bearing back-office work. These three are where the practice points first — and the named processes are the argument, not the sector labels.
Recruitment & staffing
High document volume and the same judgement applied hundreds of times a week — the shape automation likes best.
- CV parsing into the ATS, without re-keying
- Candidate screening against the brief, with reasons attached
- Job-description generation that matches the client’s voice
- ATS sync, so the same candidate never lives in three states
B2B professional services
Accounting, legal operations and insurance broking are regulated enough that a DIY prototype is a liability — the work needs evaluation suites and an audit trail.
- Client intake automation, from first email to opened matter
- Document extraction — the values inside PDFs, into the systems that need them
- Proposal generation from precedent, for the 80% that repeats
- Reconciliation and review queues that surface only the exceptions
D2C & e-commerce operations
Order volume turns small per-order tasks into whole roles. The processes are countable, which makes them quotable.
- Order-status deflection, answered from the systems of record
- Returns triage — policy applied consistently, exceptions escalated
- Supplier email parsing into structured updates
- Inventory and order data moved between store, warehouse and accounts
- Three people spend their mornings copying data between two systems.
- We lose leads that come in after 7pm.
- Every proposal takes four hours to write, and they are 80% identical.
- Our support inbox gets 200 repeat questions a week.
The industry matters less than the shape of the work: a process that can be described in one sentence and counted in hours per week can be scoped, whatever the sector. If it cannot be counted, it is not a build yet — it is a diagnostic.
FAQ
What buyers ask about scope, ownership and cost
You can buy it directly. A support professional who wants to close three times as many tickets without dropping quality, an accountant working through reconciliations, a lawyer summarising discovery, a recruiter writing outreach that does not read as machine-written — each of those is a legitimate engagement on its own terms, with no employer involved. The intake is identical to the enterprise version: your real work, your real files, a roadmap built for how you actually operate. Scope, format and timing are agreed in writing before anything begins, and the sessions are yours alone. Nothing you did not know at the start is reported to anyone, because there is nobody in the arrangement to report it to.
Sits inside your operation for a month or more and builds the AI layer for the work as it actually happens, rather than as the process document describes it. The early weeks are spent watching: which spreadsheet is really the system of record, which approval is a formality, where the same value gets re-keyed into three systems. What follows is a build — one component’s automation, a whole product’s automation, or end-to-end AI solution design and delivery — in your infrastructure, in your version control, with the same evaluation suites, failure paths, observability and runbooks as any other engagement. The difference is proximity, not standard: decisions that would take a week of email get made at the desk where the work is done.
Either. On-site placement is available in Bengaluru and elsewhere by arrangement, quoted with travel and subject to your own access, security and background-check requirements. Remote placement means working your hours rather than overlapping them — present in the same channels, the same queues and the same daily rhythm for the working day, for a defined term agreed in advance. A forward deployed engineer is placed with you against a written brief and a named client-side owner, and the term is fixed before it starts. In both forms the constraints are unchanged: credentials stay in your secret manager, code lands in your repositories, and the delegated access is revocable in one click.
Progress at the level of the organisation, never a transcript of anyone’s session. The sponsor receives the agreed objectives, roadmap coverage across the people named in the engagement, completion against the written schedule, the themes that recurred — the tool nobody knew existed, the policy everyone was quietly working around — and the automation candidates the intakes surfaced. What the sponsor does not receive is who asked which basic question, who needed a second session, or who was slowest to start. That boundary is stated to everyone before their intake, because confidentiality is what makes an intake honest, and honest intakes are the only ones worth building a roadmap from.
The intakes are an unusually accurate automation diagnostic, and they feed straight into it. Sitting with people individually and watching them work surfaces the same task done a dozen different ways, the export nobody automated, and the three steps every one of them repeats by hand every morning — which is precisely the input a Workflow Automation Sprint or an AI Agent Build needs. Where something turns out to be worth building rather than teaching, it is written up as a ranked candidate with an hours-saved estimate and the assumptions stated, so you can challenge it. The two are quoted separately and neither is a condition of the other. Teaching a person to do a task faster is the wrong answer when the task should never have reached them.
You do, on full payment. The workflows, the prompts, the evaluation suite, the runbook and the source all live in your repositories and your infrastructure from day one. I hold delegated admin access during the engagement, which you can revoke at any time. There is no vendor lock-in to unwind, because there is no vendor tenancy to be locked into.
Everything runs in your own tenancy — your cloud account, your automation instance, your vector database, your model API keys. That is deliberate on three grounds: you never lose access to your own operations if we stop working together, you see the real running cost directly rather than through my markup, and it keeps the arrangement clean under the licences of the tools involved.
Model and third-party SaaS costs are billed to your own accounts and are never resold through me — I do not want to carry token-price risk on your behalf and you should not pay a margin on it. Every AI build ships with a cost-per-conversation or cost-per-run projection so the monthly number is known before launch rather than discovered afterwards. Ongoing support is optional, quoted monthly against the tier you choose, and never bundled into the build price.
Who on your team should be better at this by next quarter?
Bring one person, one team, or one process that is costing you hours every week. You will have a reply within 24 hours — and if it is not something I can help with, I will say so rather than sending a proposal.
Prefer email? hello@soloz.ai