“We win more work than we can quote.”
Estimators retype specifications and chase supplier prices, so quotes take days. The customer who needed a price by Thursday buys elsewhere — lost orders, not lost hours.
Take on more orders without adding the same overhead — and own the system that does it.
We find the work that caps your orders or your margin, redesign it with your people, and put AI employees on your org chart to do it. You pay a fixed fee for the work and a success fee only when it hits figures you signed off. When we leave, it keeps running on your servers, under a perpetual licence.
Talk through one constraint30 minutes with the partner accountable for your project
See the diagnostic — from €7,500
Team fromBCG · McKinsey · Microsoft · Meta · Maersk
Live in a UK investment firm18 AI employees on its own serverdraft
Delivered70% of a reconciliation workload automated
The problem
Your ceiling is usually inside the company, not in the market
“Every new contract means another hire.”
Order changes, delivery dates and project hand-offs are chased by hand across inboxes and spreadsheets. Payroll grows with revenue; margin doesn’t.
“Nothing important moves until I’ve seen it.”
Discount exceptions and supplier disputes land on your phone. Decisions queue behind one calendar — yours.
“If I stepped back, what would a buyer actually get?”
How jobs are costed lives in two or three heads. A buyer prices in that risk; a successor inherits phone numbers, not a system.
Typical ground: B2B distribution, engineering and professional services, hospitality operators — companies of 50–500 people where expensive specialists spend their days on information work that sits on the path to revenue.
Results
Delivered, not projected
70%
Financial services · 20+ partner network
70% of the reconciliation work automated; 5× fewer errors, fixes 3.5× faster.
120→20
Large technology company · 120+ reports
120 reports consolidated to 20; more than $5M a year in running costs saved.
$7M+
European airline · 100+ aircraft
$7M+ saved a year; 4.5% off repair spend.
Approach
Your people design the new work. We stay until it runs.
The people who do the work today map it with us in the first week — a process your team did not help design is a process your team will route around.
- 1Diagnosewhere operations cap your orders, your margin or your own time, measured in your numbers.
- 2Decidewhich three to five changes are worth the money, which are not, and the baseline each will be judged on.
- 3Redesignwhat your people do, what AI employees do, what your existing systems keep doing.
- 4BuildAI employees are configured on Orograph, our platform, from components already running in other businesses — set up, not coded from scratch — inside your infrastructure.
- 5Run & improveresults reported against the baseline; AI employees adjusted as the business changes.
Under all five: change management from day one — your team maps today’s work, tests the new work on real cases, and learns on its own quotes and orders, not on a demo.
Where to start
Start with a diagnostic you can use even if you stop there
OfferBusiness Value & AI Opportunity Diagnostic
Pricefrom €7,500
FeeFixed · 1–2 weeks
If you go aheadCredited in full against the first wave
We work from your figures and with the people who run the work, not from a questionnaire.
The report is yours — act on it in-house, or take it to another firm. And if the honest answer is that AI is not your best investment here — a pricing change, a hire or an off-the-shelf tool would do more — the diagnostic says so.
You receive
- A map of the constraints that cost you most — in orders, margin, cash or owner time
- 3–5 priorities, ranked by economic effect
- A preliminary financial model for each, including what it would cost to run
- The risks: data, people, customers, control
- A first-wave plan: scope, acceptance stages and the figures each stage will be paid against
How we are paid
Part of our fee is paid only when the figures you signed off move
OfferFirst wave · one contract, split into acceptance stages
Fixed feefrom €60,000
PlusA success fee
Duration8–12 weeks
Before we build anything, we record today’s figures with you — time per quote, orders handled per person, error rate, conversion — and you sign them off. The fixed fee pays for the work as it happens; the success fee is due only when those figures are met. You decide go or no-go after the diagnostic, and you can stop at the end of any stage.
How we count the money (illustrative)
Your estimators send 40 quotes a week and win one in four. If the first wave cuts quote time from three days to one, the value is not “hours saved” — it is the extra quotes your team can now send, at your real win rate and gross margin, minus what the system costs to run. Give a team back 500 hours and you have capacity, not 500 hours off the payroll; we only count it when it turns into orders or a hire you no longer need.
And your people?
The aim is that the team you have handles more work — not that it shrinks. Who does what changes; the people who know your customers stay at the centre of it.
Ownership
We stay until it works. You keep it when we leave.
A consultant leaves you a strategy. A software house leaves you a system. A SaaS vendor leaves you a login. We leave you a working capability — and the right to keep running it.
Your infrastructure.
The platform and your data run on your servers or your own cloud account.
Your choice of model.
Any major model provider; local models deployed on request (some tasks lose a little quality on local models — we measure it with you). When a better or cheaper model appears, you change a setting, not your agents.
A perpetual right to use, included in the project.
If we part ways, the system keeps running. The right to use Orograph is yours for good; its code stays ours; your data stays yours.
Free standard updates. Paid operations, if you want them.
New releases cost nothing. Installing them, monitoring, an SLA and support are a separate service — from us, your own team or your IT partner.
One honest caveat: when an AI employee uses an external model provider, the text it sends to that model leaves your perimeter. If that matters, run a local model and it does not.
We want to stay your partner because we keep adding value — not because you cannot leave.
Control
Not a briefing pack. A seat in the company.
Elsewhere, an AI agent is a contractor with a visitor pass and a briefing folder: the pass in one system, the folder in another, neither aware of the other, the folder printed once.
With us, an AI employee sits on your org chart. It has an area of responsibility, a manager and limits on what it may do. Every permission carries who granted it and when. Its knowledge of your company works like a ledger, not a wiki: each entry has a source and a date, a correction is a new entry, and a person’s word outranks an agent’s.
| Your question | Typical setup today | With Lakestrom |
|---|---|---|
| Who looks after customer X? | People in the CRM, agents in a separate list | One answer: a person or an AI employee, since when, assigned by whom |
| Move a customer from an AI employee to a manager | Reconfigure, brief people, update documents | One change; both sides see it; “who handled it in September?” still has an answer |
| Why did it do that? | Logs, permissions and data in three places | One chain: action → the fact it relied on → the permission → who granted it |
| A new AI employee starts | Someone writes it a briefing | It starts with its area, its customers and a current company brief |
| Where are the gaps? | Invisible across separate systems | A process with no owner, a customer with no lead, an overloaded AI employee — on one chart |
Economics
Why bespoke costs less here — and who runs it afterwards
Most of the system already exists before your project starts. Orograph, our own platform, changes the economics:
An AI employee is a job description, not a program
Duties, tools and limits are settings, so a new one is set up, not coded.
Tools are built once and shared
The tool that reads a supplier’s quotation is reused by every AI employee that needs it; improve it once, all of them improve.
Every change is a new version
You see what changed and can go back.
Every action is on record
What it read, decided and did; quality is checked in one place.
After the wave, you choose who runs it — and can switch without changing platform.
| You run it | We run it for you (Managed AI Workforce) |
|---|---|
| Platform, AI employees and know-how stay with your team. We remain your platform supplier, with support if you want it. | Platform and data stay with you. We manage your AI employees within limits you approve: duties, permissions and KPIs, quality and exceptions, cost and model choice, development, a monthly report. |
What you always keep: the goals, approval of every permission, your systems and data, and the right to take it in-house on the same platform. Managing the AI employees is separate from operating the platform; model usage is agreed separately and passed through at cost.
Pricing
What it costs — and exactly what you get
Internal draft · proposed price points for discussion · all prices excl. VAT
We charge in two parts: a fixed fee for the work (paid as it happens — we don’t work purely on results) and a success fee tied to figures you sign off before we build. After go-live, a monthly running cost covers whatever you ask us to keep running.
Diagnostic
from €7,500fixed
You get
- Interviews with you and 4–8 people who run the work; a read of your figures (orders, quotes, margin, headcount by process)
- A constraints map: where operations cap orders, margin, cash or your own time — each with an estimated annual cost
- 3–5 priorities ranked by economic effect, each with a preliminary financial model (value, build cost, running cost, payback)
- A risk read: data, people, customers, control
- A first-wave plan: scope, acceptance stages, the baseline figures and the proposed success-fee formula
- A 90-minute decision session with you
Paid: 100% at start. Credited in full against the first wave if you go ahead.
First wave
from €60,000fixed + success fee
You get
- 1–3 processes redesigned with the people who run them
- AI employees built and live on your infrastructure — typically 3–8, each with its duties, limits, manager and permission record on your org chart
- Integration with the systems the work runs through (email, CRM/ERP, documents, spreadsheets)
- Testing on your real cases before go-live; approval rules for anything risky
- Your team trained on its own quotes and orders; written handover and runbook
- The Orograph licence: perpetual right to use, free standard updates
- 3 months of operations included after go-live (updates, monitoring, support)
Paid: fixed fee 40% at start, 30% at mid-wave acceptance, 30% at go-live.
Success fee: up to 25% of the fixed fee, due when the signed baseline figures are met — measured over the first 90 days after go-live.
Strategy & transformation programme
from €25,000fixed
For owners who want the whole map before the first wave: target operating model, business cases for each area, a 12–18-month roadmap, the organisation and skills you will need. Can run alongside the diagnostic; credited against later waves at 50%.
4Running cost after go-live — choose one
| Item | Platform care | Managed AI Workforce |
|---|---|---|
| Price | €3,000 / month | from €6,000 / monthup to 10 AI employees; +€500 per additional |
| Updates installed, monitoring, backups | ✓ | ✓ |
| Support | Business hours, next-day response | Business hours, same-day response; critical issues 4 h |
| Running the AI employees day to day (exceptions, quality checks) | — your team | ✓ |
| Tuning cost and model choice | — | ✓ quarterly |
| Developing them as the business changes | — | ✓ up to 2 days a month included |
| Monthly report: results, quality, cost | — | ✓ |
| Success fee | — | optional: share of measured improvement on agreed KPIs, in exchange for a lower base |
Model usage (the AI providers’ charges) is billed at cost, or paid directly by you to your provider. With local models it is your own hardware cost.
5Optional
Change management retainerfrom €4,000 / monthCoaching managers, onboarding new staff into the new way of working.
Extra development€1,200 per day
What stays free, for good
The perpetual right to use Orograph, standard updates, your data, your AI employees’ definitions and history. You can stop paying us anything and keep running the system.
Proof
It already runs inside a live business
A UK investment firm
18 AI employees on the firm’s own dedicated server, 862 pieces of work handled in 90 daysdraft. We operate it for them; every intervention we make is logged.
18AI employees on the firm’s own serverdraft
862pieces of work handled in 90 daysdraft
Reconciliation across 20+ partners
Financial services
Bank, broker and platform data arrived as CSV, Excel, PDF and API feeds and was reconciled by hand. We built one mapped data platform; AI agents flag mismatches and fix them once a person approves.
70% of the work automated · the same volume now takes a third of the back-office effort · 5× fewer errors, fixes 3.5× faster · 99.99% data accuracy.
Analytics over 120+ reports
Large technology company
The central analytics team had become the bottleneck. One consolidated hub plus an assistant that sends people to the right data.
120 → 20 reports · >$5M a year in running costs saved.
Lead ranking
One large AI prompt ranked every sales lead — slow, expensive, mistakes hard to trace. Split into narrow specialist agents:
ranking 72% faster at 54% lower cost, with a fifth fewer errors.
Team
You will work with the people who sign the proposal
BCG·McKinsey·Microsoft·Meta·Maersk·Tuck School of Business
Sergey Podolskiy
Co-founder & Partner; accountable for your project.
Led the redesign of Maersk’s technology organisation — operating model, governance, legacy modernisation. Former COO (operations in 70+ countries) and CRO. Ex-BCG, Tuck MBA. Founded Orograph; leads Lakestrom’s team of ex-Microsoft data & AI engineers.
Eugene Sidorin
AI economics & operating finance.
Director of Finance, AI & Infrastructure at Meta. Seven years at Microsoft running growth, profitability and cloud margin for Business Applications; former CFO of a $60M-backed AI start-up. Tuck MBA.
Dima Verbitckii
Pricing & commercial performance.
Associate Director, Commercial Performance & Pricing at Kao. Previously revenue-growth management and pricing at Pandora; consultant at BCG and McKinsey.
Every project has a second named person who knows it as well as the lead. Current employers are shown for background; they are not party to Lakestrom engagements.
Fair questions
Doubts you should have, and the terms that answer them
“They don’t understand my business.”
We start with your numbers and how the work really runs — quotes, orders, who approves what — not with a menu of AI use cases.
“I’ll get a pretty deck.”
The diagnostic ends in decisions and a financial model, and it may conclude that AI is not your best investment.
“A partner sells, juniors deliver.”
The partner in the first conversation is accountable through delivery, and the team is named in the proposal.
“The project will run away from me.”
Fixed stages with acceptance criteria; a success fee that is only due when your signed figures move; go/no-go after the diagnostic; stop at the end of any stage.
“What if it does something wrong with a customer?”
Nothing risky goes out without a person: you approve every permission, AI employees are tested on your real cases before go-live, behaviour is versioned, and any change can be rolled back.
“They’ll push their own software.”
Our rule: we recommend Orograph only where a bespoke build pays for itself. Otherwise we recommend an off-the-shelf product — or no automation at all.
“My people will resist.”
They help design the new work from the first week and learn on their own quotes and orders.
“You’re a small firm. Who supports us in three years?”
Fair question — we are small. So the setup doesn’t depend on us: it runs on your infrastructure under a perpetual licence, with a second named person on every project, a written handover and documentation, and the option to run it in-house. If we run it for you, that is a separate service you can end without losing the system.
Contact
Start with one constraint, not an AI strategy
Bring the one thing that most holds your business back: quotes that take too long, orders that pass through too many hands, decisions that wait for you. You’ll discuss it with the partner who would be accountable for fixing it.
In 30 minutes we check four things:
- 01what the problem costs
- 02how much of it runs through you
- 03how ready your people are
- 04whether the data can be reached
We won’t open with our platform. You leave knowing whether a diagnostic is worth your money.
Probably not a fit if you want a cheap chatbot, your constraint is purely physical work, nobody senior will own the change, or the data cannot be reached.