ZIAVE

System Blueprint · Full build on modeled data · Not client work

The service business drowning in manual follow-up

What we'd build for an expert services company whose growth is capped not by demand but by the founder's inbox. Every number below is a labeled assumption; the architecture and the failure modes are the real content.

Scope: Automation · CRM · Lifecycle · Retention · Published 2026-07-12

01 · The archetype

Who this is about

An agency, consultancy, or expert practice that's genuinely good: referrals arrive, the site brings inquiries, clients come back. And everything between those events is manual. Follow-up lives in the founder's head. Proposals take days of evenings. Past clients are forgotten the moment a project closes. The team does robot work between the billable hours, and the robot work is winning.

Modeled, not measured

Modeled profile: ~40 inquiries per month across referrals, site and LinkedIn; average project ~€5k; a team of 3 to 10 where the founder still touches every deal. Swap in your numbers; the shape of the problem survives the substitution.

02 · The presenting problem

“We need more leads”

The request is more leads. The audit almost always says: you have the leads, you're burning them. A high-touch sale is won or lost in the hours after the inquiry, and lost again in the days a proposal takes to arrive. The leak map ranks like this:

Leak map: modeled ranking
LeakWhat it looks likeWhy it ranks here
1. Response latency and dropped threadsInquiries answered in days, follow-ups sent when someone remembers, warm threads that simply stop.It sits directly on revenue, costs nothing to fix relative to its size, and compounds with every other fix.
2. Proposal cycle timeEvery proposal is hand-assembled from the last one. Days pass; the client's urgency cools; scope discussions restart.Speed is a proxy for competence in professional services. The delay taxes the close rate.
3. The dormant client baseDozens of past clients who liked the work and were never contacted again. No check-ins, no review asks, no referral prompts.It's the cheapest revenue the business owns, and it's currently priced at zero.

03 · The system

What we'd build

The rule: humans handle judgment and relationships, software handles motion and memory. AI enters only where it drafts and a human approves. Nothing customer-facing goes out unreviewed.

Target architecture: from inquiry to repeat client
  1. One pipeline, all sources

    Referrals, site forms, LinkedIn DMs and email all land in one CRM pipeline with enforced stages. No lead lives only in an inbox again.

  2. Instant intake with AI triage

    Every inquiry is enriched and qualified on arrival; the AI drafts a tailored first reply and a human approves it. Response time drops from days to minutes without a single robotic message.

  3. Follow-up sequences with human checkpoints

    Threads can't silently die: the system schedules the next touch, drafts it, and asks. Judgment stays human, persistence becomes mechanical.

  4. Proposal assembly

    Scope blocks, pricing blocks and case blocks, assembled per deal in hours. The founder edits instead of writing from zero.

  5. Lifecycle and referral loop

    Post-project check-ins, review requests and referral prompts run on schedule, in the founder's voice, with value attached. The client base becomes an asset with a pulse.

  6. The operations dashboard

    Response time, pipeline by stage, proposal cycle time, reactivation revenue. One screen, one weekly review.

Example stack · illustrative · not a recommendation

Pipedrive · Make · Slack · Google Workspace · Looker Studio

One way this architecture could be implemented. The constraint picks the tools, not the other way around. No vendor affiliations.

04 · The math

The metric model

Baseline: 40 inquiries a month, a quarter become calls (slow responses and dropped threads eat the rest), 40% of calls close. Four projects at €5k: €20k a month. The model changes nothing about demand.

Inquiry-to-revenue: baseline vs modeled (all inputs assumed)
StepBaselineAfter fix 1: intake + follow-upAfter fix 2: + lifecycle loop
Inquiries / month404040
Inquiry → call25% → 10 calls45% → 18 calls (minutes, not days; threads can't die)45% → 18 calls
Call → closed40% → 4 projects40% → 7.2 projects40% → 7.2 projects
Reactivated past clients00+2 projects / month (modeled from a dormant base of 100+)
Revenue / month @ €5k€20k€36k (+80%)€46k (+130%)

Modeled, not measured

Every input is an assumption. The load-bearing one is inquiry-to-call rising from 25% to 45% when response time drops to minutes and no thread is ever silently dropped; the rest is arithmetic. Replace the inputs with your pipeline's real numbers and the output is your bottleneck cost. That's the growth diagnostic, live.

05 · The sequence

How it ships

Build sequence: working increments, instrumented from day one
  1. Increment 1: the pipeline

    One CRM, all sources connected, stages with entry criteria, current deals migrated. The inbox retires as a database.

  2. Increment 2: intake + follow-up automation

    AI triage with draft-and-approve, follow-up scheduling with human checkpoints. This is where the response-time metric moves.

  3. Increment 3: proposal assembly

    Block library built from the last ten winning proposals, assembly flow per deal. Cycle time drops from days to hours.

  4. Increment 4: lifecycle + referral loop

    Check-in, review and referral flows over the historical client base, warmed up gradually and measured as its own revenue line.

06 · Failure modes

What we'd get wrong first

Known failure modes and their mitigations
FailureMitigation
AI answers a client unsupervisedNever in this build. Draft-first, human-approve is a hard gate, not a setting. Trust in a services business is the product.
Automation reads as robotic to high-touch clientsSequences are written in the founder's voice from real past emails, and every checkpoint is a human decision. Clients should feel more attended to, not less.
Proposals become boilerplateBlocks are raw material, not the proposal. The founder's judgment assembles them per deal; the system just kills the blank page.
Reactivation reads as spamEvery reactivation touch leads with something useful: an insight, an audit observation, a relevant change. “Any projects for us?” is banned copy.

This blueprint is a demonstration: modeled inputs, real architecture, real sequence, real failure modes. It's published so you can judge how we think before you pay for it. When a client system in this shape ships and produces measured results, a real case replaces it. Same structure, real numbers. Published under our publication standard.

All case studies

The dispatch

Get the next one when it ships

We publish one or two pieces a month: teardowns, blueprints, leak entries. When something ships, you get it. No weekly filler, no drip funnel, unsubscribe anytime.

The first step

See where your system leaks.

One 30-minute growth diagnostic on your actual funnel: the constraints slowing growth and the opportunities with the highest leverage, ranked by what they cost you, whether the fix is positioning, CRM or AI automation. You leave with the one-page Growth Leverage Map and a prioritized fix sequence, whether or not we ever talk again.

Get your Growth Leverage Map

30 min · Free · No pitch · the map is yours either way

Not ready for a call? Get your Growth Score in 3 minutes and see where your system stands first.