Operations

Building the Tiding BD Stack

From manual outreach to systematic pipeline generation. What happens when you treat business development as infrastructure, not improvisation.

Recruitment agencies spend 60-70% of senior consultant time on business development. Not interviewing. Not negotiating offers. Not building relationships with placed candidates. Business development.

And here's the thing: most of that BD is reactive. A job board alert pops up. A referral comes in. Someone remembers to follow up on that conversation from three months ago. Maybe.

The gap is obvious: systematic, repeatable pipeline generation that doesn't rely on a human remembering to follow up. The thesis is simple — what if BD was infrastructure, not improvisation?

That's what we're building at Tiding. And we're doing it in public.

The Stack Architecture

The Tiding BD stack is designed around a simple principle: Zac (the AI) does the research, drafting, and systematic follow-up. Human oversight gates are in place while Zac trains — monitoring quality, reviewing drafts before they go out, handling the conversations that require judgment. Once validated, those gates open. The oversight stays, but the friction goes.

Here's how it flows:

Input Layer

XML Job Feeds  •  LinkedIn Sales Nav  •  Trade Directories  •  Companies House  •  Manual Research  •  CRM Imports

Enrichment Layer

Apollo (contacts)  •  Email Verification  •  Headcount Data  •  ICP Scoring  •  Exclusion Checks  •  Tiering

Decision Engine (MPC Query)

Candidate Profile → Market Map → Contact Discovery → Exclusion Check → Draft Generation → Queue for Review

Execution Layer

Cold Email  •  WhatsApp Business  •  LinkedIn DMs  •  Follow-up Sequences

Pipeline & CRM Layer

CRM CRM  •  Discord Alerts  •  booking system Booking  •  Dashboard  •  Reporting  •  Handoff Protocol

Design principles: Modular (each layer swappable), API-first (everything programmable), Human-in-the-loop (AI drafts, human approves). The stack doesn't replace the recruiter. It removes the administrative overhead that stops recruiters from recruiting.

What Works (So Far)

XML Feed Processing

We processed a single feed containing 1,927 job postings. From that:

1,927
Jobs Processed
942
Unique Companies
609
Tier 1 Signals
59
CRM Matches

The classification happens automatically: Tier 1 (respond within 24h), Tier 2 (48h), Tier 3 (weekly batch). No more scanning job boards manually.

CRM Enrichment

We took a "clean" CRM of 69 companies and discovered it was full of noise. Fireplaces. Fireworks retailers. Hockey clubs. All caught by keyword-matching "fire" without context.

After a reset:

Data quality beats data quantity. 38 clean, segmented records > 69 messy ones.

MPC Query v2

We rebuilt our Most Placeable Candidate process from scratch. The old version produced 15 empty shell drafts — no contacts, identical subject lines, copy-pasted briefs in every field.

The new version:

First run produced 10 complete drafts with contacts, subjects, and email copy. Ready for exclusion check and send.

Infrastructure

The Bolt-On Discovery

Here's something we didn't anticipate: this stack doesn't have to be Tiding-only.

We're currently running a trial with another agency in a market we identified as sub-optimal for Tiding to build a brand in, but where systematic BD would be valuable. Education and apprenticeship recruitment — proven demand, good fees, but not where we want to focus our brand-building energy.

The experiment is testing whether Tiding BD can layer over existing infrastructure:

Early signals are promising. First week produced 15 MPC drafts. XML feed integration is live. The stack is proving it can adapt to different schemas and workflows.

This could be a second business line: BD-as-a-Service for agencies that want systematic pipeline generation without building it themselves.

The Challenges (Real Talk)

Data Quality at Scale

The "clean" CRM wasn't clean. The XML feed contained noise. Every data source requires validation and enrichment. Getting data to requisite quality is harder than getting the data in the first place.

Scale vs. Precision

We can process 2,000 jobs in minutes. But identifying which 50 companies are worth targeting? That still requires judgment. We're building heuristics (company size, hiring velocity, role seniority) but the balance between casting wide and targeting tight is still being calibrated.

The "Good Enough" Trap

There's a tension between perfect data and shipped work. We've leaned toward shipped > perfect, but that means living with incomplete records, manual fallbacks when APIs hit limits, and documented workarounds. It's functional but not elegant. Yet.

What's not solved yet:

Work in Progress

This is infrastructure building. It's unglamorous. It doesn't generate placements directly. But without it, every placement relies on luck and memory.

The current operating cadence:

We're still tweaking to get this perfect. The MPC Query works but needs more test runs. The enrichment layer needs more data sources. The execution layer needs approval workflows, not just drafts.

What's Next

The stack is the product. The placements are the proof.

Tiding is an AI-first recruitment company. We build infrastructure for systematic business development, then apply it to markets with persistent skills shortages. This post is part of our build-in-public series — real metrics, real failures, real progress.

← Back to Notes