The operating layer
behind predictable revenue.
Most mid-market companies do not have a sales problem. They have a systems problem — a CRM nobody trusts, ten tools that do not talk to each other, and people doing by hand what software should do by default. We engineer the layer that fixes it: CRM architecture, automation, and AI that runs in production.
- Who we build for
- B2B companies with 20–500 employees and a real sales team
- Core stack
- Close CRM · Make.com · n8n · Claude & GPT · Meta, LinkedIn & Google Ads
- Time to production
- First working system live in 4–6 weeks, not 4–6 quarters
- Ownership
- Every account, blueprint and document belongs to you from day one
Your growth is not limited by effort. It is limited by handoffs.
Every mid-market company we meet has the same five leaks. None of them are visible on a P&L until they have been running for two years.
The CRM is a graveyard
Reps log activity because they are told to, not because it helps them. Pipeline stages mean different things to different people, forecasts are guesses, and the one report leadership actually wants gets rebuilt in a spreadsheet every month.
Copy-paste is the integration
Lead comes in on a form, someone re-types it into the CRM, someone else re-types it into the quoting tool, and a third person re-types it into accounting. Four systems, one truth, zero synchronisation — and every retype is a chance to lose the deal.
Follow-up depends on memory
The deals that die are rarely lost to a competitor. They go quiet, nobody notices for eleven days, and by then the buying window has closed. Consistency is a system property, not a discipline problem.
Marketing spend has no closed loop
You know cost per lead. You do not know cost per qualified opportunity, per channel, per campaign, per creative. So budget gets allocated on gut feel and the channel that actually produces revenue gets underfunded.
AI got bought, not built
A licence was purchased, a pilot was run, a demo impressed someone. Nothing reached production, because nobody connected it to real data, real workflows, and a real definition of "correct". AI without an evaluation harness is a science project.
We close all five — as one system
Not five vendors, five tools and five invoices. One architecture where the CRM is the source of truth, automation moves the data, AI handles the judgement-heavy steps, and paid acquisition feeds it with attribution intact.
Our methodFour disciplines. Deliberately one team.
Each of these is a full practice in its own right — and each one gets substantially better when the same people build the other three. Below is the short version; every page goes deep.
Close CRM & Sales Systems
Migration, pipeline architecture, Smart View work queues, sequences, power dialling, custom objects and forecasting you can defend in a board meeting. We are a Close-first shop — and we will tell you honestly when Close is the wrong answer.
Read the full scopeAutomation Architecture
Make.com and n8n built like software, not like a hobby: idempotent scenarios, explicit error handling, data stores for state, blueprints in version control, alerting into Slack, and an operations budget that does not quietly triple.
Read the full scopeApplied AI Systems
Production AI, not demos. Document and email triage, retrieval over your own knowledge base, call transcripts turned into CRM fields, voice qualification agents — each shipped with an evaluation set, guardrails and a human in the loop where it matters.
Read the full scopeGrowth & Demand Generation
LinkedIn, Meta and Google campaigns designed as a two-stage funnel and wired back to closed revenue. Server-side tracking, a UTM taxonomy that survives contact with reality, and reporting on cost per opportunity — not cost per click.
Read the full scopeSix layers, one source of truth.
This is the reference architecture we start every engagement from. Yours will differ in the details — never in the shape.
How a project actually runs.
Fixed scope, fixed price per phase, and a working artefact at the end of every one. You can stop after any phase and still own something useful.
Diagnose — 1 to 2 weeks
We map the current state: every tool, every handoff, every manual step, and every place data is re-typed. We sit with reps and back office, not just management, because the workaround is always where the truth lives. Output: a written systems audit, a prioritised backlog with effort and impact per item, and a target architecture.
Design — 1 to 2 weeks
Data model, field dictionary, pipeline definitions with explicit stage exit criteria, automation flow diagrams, naming conventions and access model. Signed off before a single scenario gets built, because rework in design costs an afternoon and rework in production costs a quarter.
Build — 3 to 8 weeks
Implementation in your own accounts, in weekly increments you can see working. Migration runs against a staging copy first with a reconciliation report. Every automation ships with error handling, alerting and a documented rollback path.
Adopt — 2 to 4 weeks
The hardest part of any CRM project has nothing to do with software. Role-based training, a one-page daily operating rhythm per role, recorded walkthroughs for onboarding future hires, and adoption measured as a metric rather than assumed.
Operate — ongoing, optional
Monitoring, incident response, an iteration budget for the changes that only become obvious once real volume hits, and a quarterly review against the original backlog. Or we hand over completely and you run it. Both are fine — that is the point of documenting properly.
Where we are genuinely strong
- Sales-led B2B with a real pipeline. Considered purchase, multiple stakeholders, a cycle measured in weeks or months, deal values that justify engineering the process.
- Service businesses that run projects. Where a signed deal becomes a delivery workflow, and the handoff between sales and operations is where things currently break.
- Recruiting, staffing and professional services. Two-sided pipelines — clients and candidates — with heavy document and communication load. Automation and AI pay back fastest here.
- Manufacturers and technical suppliers going digital. Excellent product, strong reputation, and a commercial process still running on email, PDFs and one person's head.
- Companies with a German or European client base. We build for GDPR, consent and documentation requirements as a design constraint, not a retrofit.
Where we are the wrong partner
- Pure self-serve or low-ticket e-commerce. No human sales motion means most of what we are good at does not apply.
- Enterprises committed to Salesforce or Dynamics with a large internal team. You need a platform-specific partner with a bench, not us.
- "Just make the tool work by Friday." If nobody is available to make decisions about process, no implementation survives the first month.
- Automation as a headcount-reduction promise. We automate work, not people. Every project we have seen succeed reallocated capacity — it did not simply remove it.
- Anyone looking for the cheapest quote. We are not expensive, but we are not the cheapest, and the difference is usually documentation and error handling — the two things that decide whether it still works next year.
What is actually different here.
We are operators, not resellers
We run these systems in our own businesses before we recommend them. No referral-fee-driven tool selection, no certification badges determining the architecture. If a free tool fits better than a paid one, we will say so.
Everything is documented to hand over
Field dictionary, flow diagrams, runbooks, recorded walkthroughs, blueprints exported and version-controlled. The test we hold ourselves to: could a competent new hire take this over in a week without calling us? If not, it is not finished.
You own every account
The Close workspace, the Make organisation, the ad accounts, the API keys, the domain, the code. Registered to your company, billed to your card, administered by your people. We work inside your tenancy — never behind an agency wrapper.
Compliance is built in, not bolted on
For German and EU clients that means lawful bases documented per processing activity, consent captured where it is required, transfer safeguards in place, and outbound designed around what UWG and GDPR actually permit — before campaigns go live.
Error handling is not optional
Anyone can build a happy path in an afternoon. What separates a demo from a system is what happens when the API returns 429, when a field is empty, when a duplicate arrives, and when someone changes a dropdown value. That is where our time goes.
One point of contact, senior throughout
The person who scopes your project is the person who builds it. No account manager relaying requirements to an offshore team, no junior learning on your budget, no context lost between the call and the ticket.
Before you get in touch
What does a project cost?
Diagnose is a fixed fee and deliberately small — it is a two-week engagement that produces a written audit and target architecture you can act on with or without us. Design and Build are quoted as fixed-scope phases once we know what we are dealing with; there are no hourly surprises. Optional managed operations is a monthly retainer sized to your automation volume and change rate.
Most implementations for a 20–150 person company land in the range of a mid-sized internal project — meaningfully less than one full-time hire for the year, and roughly what the same company loses annually to manual re-entry and dropped follow-ups. We give you a real number after the first call, not a range on a website.
Do we have to use Close CRM?
No. Close is our default recommendation for outbound-heavy mid-market sales teams because it puts calling, email, sequences and pipeline in one place without per-seat cost sprawl, and its API is genuinely good to build on. That makes it a strong fit for maybe seven in ten of the companies who approach us.
For the rest — heavy field service, complex quoting, product-led motions, or an existing Salesforce or HubSpot investment your organisation is committed to — we build on what you have. The architecture, the automation layer and the AI work are all platform-independent.
How long until we see something working?
The systems audit lands within two weeks of kickoff. The first automations that remove real manual work are typically live in weeks four to six. A complete CRM migration with pipelines, sequences, dashboards and trained users usually completes in eight to twelve weeks depending on data quality and how many systems need to be connected.
The single biggest variable is not our capacity — it is how quickly your side can make decisions about process. We tell you exactly which decisions are needed and when, at the start.
We are in Germany. Does working with a UAE company create problems?
It is a normal cross-border services arrangement and it is handled contractually. Two things matter in practice. First, VAT: for B2B services to a German business, the reverse-charge mechanism generally applies, so you self-account for VAT and we invoice without it — your tax adviser will confirm for your specific situation. Second, data protection: where we process personal data on your behalf we act as a processor under a Data Processing Agreement, and because the UAE has no EU adequacy decision, transfers are covered by Standard Contractual Clauses with a documented transfer impact assessment.
In most engagements we work inside your tenancy — your Close workspace, your Make organisation, your cloud accounts — which keeps the data resident where you already chose to put it and keeps the processing footprint small by design.
Can you work with our existing agency or IT team?
Yes, and it is common. We usually take the systems and data layer while an existing agency keeps creative and media, or while internal IT keeps infrastructure and security. What we ask for is a clear boundary of responsibility and one decision-maker on your side — shared ownership of the same automation with no named owner is the single most reliable way to break it.
What happens if we stop working with you?
Nothing breaks. You already own the accounts and the licences are already in your name. We export blueprints, hand over documentation and runbooks, walk your team or your next partner through the architecture, and remove our access. There is no proprietary middleware, no agency-hosted connector and no scripts that only run on our machines. That is a deliberate design constraint, not a courtesy.
Do you use AI in your own delivery?
Extensively — for code generation, data mapping, documentation and analysis. It is why a two-person senior team ships what used to take five people. We are equally clear about where it does not belong: architecture decisions, anything touching client personal data without a signed agreement in place, and anything that goes to a customer without a human reading it first.
A 30-minute call, and you leave with a plan.
No deck, no discovery theatre. Tell us where the process breaks; we will tell you what we would build, in what order, and roughly what it costs — whether or not we end up doing it.