The workflow specification layer
From observed work to implementation-ready specifications.
Before any team can implement a new system — or an agent that automates the old one — someone has to write down how the work actually happens. That's where transformation projects stall. Aperture observes the work directly — across browsers, desktops, and your systems — and produces specs your implementation team can build against.
Observes work across
What Aperture does
Raw work becomes an engineer-ready specification.
Forty thousand low-level events per week become a few dozen workflow specifications — each one a contract your team can build against, whether the target is an agent, an ERP rollout, or a process redesign.
The gap
The discovery problem nobody solves
Standard operating procedures don't match how the work happens today. They were accurate the day they were written and have drifted ever since — through reorganizations, system migrations, exception handling, and the quiet workarounds people invent to get the job done.
Process mining dashboards don't ship code. They produce charts that have to be translated, by hand, into an implementation. By the time the chart becomes code, the chart is months old and the underlying process has moved on.
Whether the work is being done by your engineers or a delivery partner, the discovery phase is where weeks disappear. Multiply that by every workflow you want to automate. Aperture compresses six weeks of discovery into three days of structured specs.
Why this matters
Discovery is where transformation projects stall.
Mapping a single workflow by hand takes weeks. Multiply that by every workflow you want to automate, and the engagement runs out of runway before code is written.
What that ratio means at portfolio scale.
Traditional discovery is serial — one SME, one workflow, one whiteboard at a time. Aperture runs in parallel across every active user the moment capture begins.
Where it lives
The missing layer between strategy and code.
Aperture replaces the weeks-long, interview-driven discovery phase that sits between deciding what to automate and writing the first line of implementation code — whether that's an agent or an ERP rollout.
You've decided what to automate. The agenda is set; partners are picked; budget is approved.
We observe the work directly across browsers, desktops, and your systems — then produce structured specs, automation scores, and eval sets. Three days, not six weeks.
Your engineers — or a delivery partner — pick up the spec on day one and start building against it. No more re-discovery.
The artifact
What we deliver
Day-one inputs for the team doing the build — yours, ours, or a delivery partner's.
Workflow specification
Structured docs: trigger, completion condition, actors, ordered steps with action, duration, exceptions, decision points, and the tribal-knowledge dependencies that nobody wrote down.
Automation scoring
Per workflow: automation fit, data availability, volume × duration, exception rate, reversibility. Plotted as 'value if automated' against 'cost to automate' so you can sequence the work.
Eval set
20–50 input/expected-output pairs derived from real captured traces. The bootstrap material your team uses to validate any implementation built against the spec — before it touches production.
The pipeline
How it works
Five stages turn raw activity into implementation-ready specifications. Each stage transforms the shape of the data — from events to instances to types to scored candidates to specs.
Browser, desktop, and system events — privacy-filtered into a uniform event store.
Activity is grouped into discrete workflow instances — one billing dispute, start to finish.
Instances cluster into workflow types with shared steps, decisions, and exception modes.
Per type: LLM fit, exception rate, reversibility, value-if-automated. Plotted as cost × value.
Top workflows get a draft implementation spec — actors, triggers, ordered steps, decision points, eval set.
Audience
Who we work with
Anyone whose job depends on understanding how work actually happens before automating it.
- 01
AI-native operators
Roll-up operatorsOperators acquiring traditional services businesses. Aperture deploys inside each acquisition during the first 90 days so a common automation playbook applies portfolio-wide.
- 02
Forward-deployed engineering teams
FDE / applied AIEngineering teams embedded inside customer organizations to ship AI workflows. Aperture's specs replace weeks of discovery before code can be written.
- 03
Strategy & transformation consultancies
McKinsey · Bain · BCG · Accenture · DeloitteFirms running AI transformations for clients. Aperture replaces the SME-interview phase of every engagement, so recommendations are grounded in observed reality, not whiteboard models.
- 04
Private equity firms driving portfolio AI
Operating partners · PortCo programsPE sponsors rolling AI across operating companies. Aperture instruments each portfolio company during onboarding so the same automation playbook compounds across the fund.
- 05
Transformation programs at corporates
Internal AI / digital transformationInternal teams running migrations and AI rollouts across their own ERP, CRM, and operational stack — without paying for repeated SI discovery on every workflow.
From the thesis
"If your team has spent weeks mapping a process before writing the first line of code, you already know the problem we're trying to solve."Read the full thesis →