Our methodology

How we run a Sympatheia engagement, the operating level.

A short, opinionated read on how we structure pods, sequence deliverables, evaluate production agents, and protect the executive cadence we install. Methodology is not a deck. It’s a way of being in the room.

Operating principles

Six rules we hold ourselves to.

Not values. Not aspirations. Rules. If we break one, we owe you an explanation in writing before the next sync.

i

Stakeholder before stack.

No agent gets deployed without a named C level sponsor whose KPI it moves. If we can’t find the sponsor, the agent isn’t worth shipping. The technical work follows the human accountability, not the other way around.

ii

Data layer first; agents last.

The model breaks if the data is stale. Phase 03 always lands Layer 02 (Data Foundation) and Layer 03 (Systems of Record) integration before any Pillar 2 or Pillar 4 agent goes near production traffic. No exceptions.

iii

Evals on every agent, every week.

Every production agent has a regression eval set that runs weekly. Drift in accuracy, latency, or stakeholder perceived quality is logged into an incident report we walk through with the executive sponsor each month.

iv

Cadence beats dashboards.

A dashboard nobody opens is worse than no dashboard. We install the cadence first, Friday CRO, Monthly CMO, Quarterly CPO, and configure dashboards to fit the meeting, not the other way around.

v

Two anchor partners, not twenty.

Phase 03 ships against two anchor partners. The temptation is always to onboard ten. We resist. Iteration speed inside a tight pilot beats coverage every time, and the wave 2 rollout from a working pilot is half the calendar of an unfocused expansion.

vi

Outcomes before invoices.

Run phase fees are anchored to outcomes we agreed on at Design. If we miss the headline KPI two quarters in a row without a documented external driver, we discount, or end the engagement with a clean handoff.

How we staff

One pod. Three motions.

Sympatheia engagements are run by a single pod that scales between phases. Same faces, growing surface area. We don’t hand you off between teams.

Assess pod · 3 people

The diagnostic pod

Three senior practitioners run the six week diagnostic. Heavy on partner program experience and partner economics; light on engineering.

Engagement Partner 1
Partner Economics Lead 1
Data & Readiness Lead 1
Design + Build pod · 5 9 people

The build pod

Scales from 5 in Design to 7 9 during Build. Adds the engineering depth to ship agents into your environment without disruption.

Engagement Partner 1
Operating Model Architect 1
Agent Engineers 2 3
Data & Integration Engineers 1 2
UX / Dashboard Designer 1
Change & Cadence Lead 1
Run pod · 3 5 people

The embedded pod

The continuous operations team. Lighter weight than Build, but the same senior practitioners. They run alongside your alliance team.

Engagement Partner 0.5
Agent Operator 1
Analytics & Reporting Lead 1
Change & Cadence Lead 0.5
Quarterly Build Lead 1
Phase 01 detail

Six weeks of the diagnostic.

A working example of how Phase 01 unfolds in practice. We sequence interviews, data work, and synthesis so the executive read out lands on a fully evidenced memo, not a directional impression.

Week 1

Mobilization & stakeholder map

Three day kickoff. Stakeholder mapping session with the engaging executive. Data access provisioning. Identification of the eight executive interviews and the data set we’ll work from.

Week 2

Executive interviews

One hour structured interviews with CRO, CMO, CPO, CCO, CFO, Partner Exec, VP Sales, and 1 2 anchor partners. Transcript synthesis into the stakeholder alignment grid.

Week 3

Revenue mix & leakage

Deep dive into the partner P&L. Eight quarter look back across partner sourced, partner influenced, and direct revenue. Leakage analysis at each lifecycle stage. Per partner P&L modeled.

Week 4

Data & AI readiness

Test of five canonical queries against your data foundation: account partner matching, partner health scoring, certification velocity, MQL attribution, forecast reconciliation. Pillar by pillar readiness scorecard built.

Week 5

Synthesis & pre read

Internal synthesis week. Memo drafted to a 30 40 page board grade artifact. Risk register, opportunity register, and phased roadmap drafted. Pre read sent to engaging executives 72 hours before read out.

Week 6

Read out & sign off

Two hour executive read out. Three structured outcomes from the room: (1) memo signed off, (2) Phase 02 scope agreed, (3) executive cadence committed. If readiness isn’t there, we agree the foundation engagement scope instead.

After

Phase 02 mobilization

If we proceed, Phase 02 mobilizes within two weeks of read out. Same engagement partner. Same operating model architect. Build pod adds engineering by week 8 of the broader engagement.

The eval framework

How we evaluate agents in production.

Every agent we ship comes with a weekly regression eval. The framework below is what we install during Phase 03 and operate continuously thereafter. It’s the artifact that keeps an "AI for partnerships" engagement from drifting into "magic box that nobody trusts."

Accuracy

Held out test set per agent, refreshed monthly with new partner data. We track precision and recall against ground truth labels, e.g., did the Co Sell Router pick the partner that won the deal, did the Health Monitor flag the account that actually went at risk.

Target >90% precision on at risk flagging; >85% partner match accuracy on Co Sell Router.

Latency

Wall clock time from input trigger to actionable output. Co Sell Router latency from deal registration to Slack room creation. Forecast Reconciliation latency from data drop to variance memo. We have explicit SLAs per agent.

Target <1 hour Co Sell deal room provisioning; <4 hours forecast variance memo.

Stakeholder quality

A monthly anonymized stakeholder survey. Seven questions per agent, clarity, actionability, accuracy felt, time saved, integration with workflow, trust, recommendation. Quantified and tracked over time.

Target Net Stakeholder Score (NSS) >+40 for any agent in production.

Safety & governance

Red team probes per agent on adversarial inputs, hallucination tests, prompt injection probes, data leak tests for PII and partner sensitive info. Run weekly; results held in the Run phase incident log.

Target Zero data leak incidents; documented failure mode register per agent.
How we contract

Three contracting models.

We’re opinionated about how to structure the commercial relationship. Below are the three structures we default to, chosen at Design once we know the scope.

Fixed scope

Phase anchored fee

A fixed fee per phase, paid at gate. Right for Phase 01 (Assess) and Phase 02 (Design), where the scope is known and the deliverable is the artifact, not the outcome.

Best for The Founding Diagnostic, standalone Design engagements, and pre funded board cycle work.
Outcome anchor

Outcome weighted Run fee

A base monthly fee anchored to a headline KPI bonus (or discount). Right for Run, where we’re carrying the outcome with you and have direct line of sight on the agents driving it.

Best for Long running Run engagements with a Run pod embedded and quarterly outcome reviews.
Hybrid

Build + Run rollover

Build phase contracted as fixed fee with a Run phase rollover written in. Right when we’ve agreed at Design to be the embedded operator post launch, gives both sides incentive to ship a Build that holds at Run.

Best for End to end engagements where the same Engagement Partner runs all four phases.
Talk methodology with a partner

If our methodology lines up with how you want to run, let’s talk.

The fastest way to know if we’re a fit is to walk through a real engagement. We’ll bring a redacted Phase 01 memo from a comparable client and walk you through how the pod operated, what we missed, and what we changed in Phase 02 as a result.

Book a methodology walk through