What is proven
The workflow has its ingredients.
Ontology, stakeholder interviews, and ROI discovery are enough to move from discovery into a production decision.
Decision memo · June 19, 2026
Process Discovery needs a production home
The discovery motion is proven. What is unresolved is where the production version lives — because the surface we choose sets how much security and trust hardening stands between us and a real customer.
#01
Why we are deciding this now
Process Discovery is no longer blocked by whether the idea makes sense. It is blocked by productionizing the stakeholder interview and evidence loop — and the surface that loop runs on is what sets the cost.
What gates it
The interview loop has to actually run.
Emails must send, replies must land and be attributed, and process maps must update from that evidence — at customer reliability.
Why the surface decides
The surface sets the trust tax.
Auth, data isolation, audit, approval gates, and compliance all follow from where this lands — and that tax, not the feature, sets the timeline.
#02
The core insight: the platform tax is a trust tax
This flips the usual intuition that the most-built option is the fastest. The feature is the cheap part, and it is the same everywhere. The expensive, variable part is the security and trust hardening — and that is the entire decision.
Read it left to right: everything left of the dashed line — the Process Discovery feature — is identical on every path. The only thing that grows is the tax to make customer data safe on that surface, and that tax, not the feature, is what separates a two-month build from a five-month one.
#03
The recommendation, and the proof
The recommendation is Beam. The path diagram shows how the pieces sequence; the scorecard shows why, criterion by criterion. The rows are about trust, safety, and fit — the build speed itself is shared, as the chart above shows.
Executive scorecard — every criterion, side by side
| Decision criterion | Standalone wedgeAllow — wedge only | Beam platformRecommended | PrismAvoid as product |
|---|---|---|---|
| Best use | Fast pilot wedge | Production product home | Internal discovery & handoff |
| Security & trust tax on top | Pilot-only controls | Inherited — already paid | +2–3 months to build from scratch |
| Time to production-ready MVP | ~1–1.5 months | ~2 months | ~4–5 months |
| Customer trust & data isolation (today) | Not built — pilot only | Tenant-isolated at the API; audit & SOC 2 posture inherited | None — local data, no RLS, no SOC 2 |
| Fit with customers & sales story | New app to introduce; weak as a story | Customers already work here; one platform story | New product to stand up; weak as a story |
| Verdict | Use only to accelerate validation | Recommended | Do not use as production path |
#04
The three paths up close
The numbers are in the scorecard above. This is the one-paragraph read on what each path actually is, and the role it should play.
~1–1.5 mo pilot
Standalone app
The fastest way to a live customer signal, bought by staying narrow. It skips the security and trust tax entirely — which is exactly why it cannot hold real customer data for long.
Best used as: a parallel accelerator to Beam, not a fork in the roadmap. Push outputs into Beam the moment the API contract exists.
Why it stays a wedge
- Good for learning, weak as a system of record — it should never hold the durable customer truth.
- No approval or audit controls, so hold off on autonomous external outreach.
- The moment it needs real auth and audit, it converges on the Beam timeline anyway.
⤢ Click to enlarge
~2 mo MVP
Beam platform
Build directly where customers already authenticate, run agents, review outputs, and trust the workspace. The trust tax is already paid — multi-tenancy, audit, approval gates, and encryption are live today — so the six-week feature is nearly all that is left.
Best used as: the production system of record. Process Discovery ends in Beam anyway — map to ROI case to agent spec — so the output lives where the next action happens.
What is left to build
- Productize native reply handling for interview loops.
- Clarify permissions and RBAC for customer process data.
- Harden connector reliability for email, calendar, Slack/Teams, and documents.
⤢ Click to enlarge
~6 wk + 2–3 mo hardening
Prism
The feature is not harder to build here. The problem is that Prism is a single-user desktop app today, so going customer-facing means productionizing Prism itself first — a desktop-to-SaaS shift, with every security and trust control built from scratch.
Best used as: the internal discovery and handoff workbench that feeds Beam — genuinely useful, just not the customer runtime.
The trust tax Prism still owes
- Per-customer data isolation and workspace permissions — none today, local storage with no row-level security.
- Immutable, exportable, per-customer audit trail — today it is local JSON only.
- SOC 2, penetration test, data residency — all required before a regulated buyer, all unbuilt.
- Approval gates and an email interview channel — the active security pass is still mid-flight.
⤢ Click to enlarge
#05
The real cost is trust
The hardening — per-customer isolation, audit trails, approval gates, encryption, compliance — is not optional polish, and skipping it does not save the time. It relocates the cost to the worst possible moment: a customer incident.
Skipping looks cheaper for a long time. That is the trap: the cost does not disappear, it compounds in the dark and surfaces as a breach or a failed enterprise review — exactly when it is most expensive and least recoverable.
Risk 1
Cross-tenant exposure and breach
Without per-customer isolation, an immutable audit trail, encryption-at-rest, and approval gates on outbound email, sensitive process evidence and stakeholder quotes are exposed to cross-tenant leakage, unreviewed external messages, and untraceable actions. On a surface that has not paid the trust tax, that is exactly the gap a breach walks through.
Risk 2
Trust you only get to lose once
Regulated buyers in banking and insurance ask for the audit trail, data residency, and SOC 2 posture before they move past POC. No hardening means no enterprise deal — and a single breach with a design partner erases the credibility the whole discovery motion depends on. Trust is the product here.
Why the surface decides everything
Beam already carries most of this hardening; Prism would have to build all of it under deadline pressure. "Just ship it in Prism" quietly means "ship the trust risk too" — and that is the one thing in this motion you cannot rebuild after it breaks.
#06
The one decision
This is a single call, not a list: which surface do we productionize Process Discovery on? Pilot scope, owners, and the success metric all follow from that one choice.
Build Process Discovery on the Beam platform.
Use the standalone app only as a short pilot wedge if we need a faster signal, and keep Prism as the internal discovery workbench. Everything else — pilot scope, owners, the success metric — is execution that follows this one decision.