Intelligent Systems & Agentic Workflows

Opportunity Intelligence

From Signal to Conversation.

The difficult idea

Recognize where a company is already moving, understand why it matters, and begin a relevant conversation—every day, across a changing market.

FTL role: Strategy, agentic workflow design, AI engineering, product design, infrastructure, validation and internal operation
Domain: Internal Systems & Business Development
Format: Agentic workflow · Opportunity intelligence platform · Internal operating system
Credits: Designed, engineered and operated internally by Faster Than Light.

Context

FTL’s best new partnerships often announce themselves indirectly: a company hires for a capability, builds a new team or publishes a role that reveals a strategic goal. Tracking those signals manually across the DACH region meant that useful context arrived late and fragmented.

Creative vision

Build an internal intelligence system that recognizes motion before it becomes a brief. Every qualified signal should become a sourced understanding of the company, a credible FTL contribution and a conversation with a specific reason to begin.

Challenge & constraints

Public job data is noisy and temporary. Companies describe similar needs in different language, while research, contact data and portfolio evidence carry different trust and privacy boundaries. The workflow had to automate repetitive investigation without automating FTL’s judgment.

What was made

A private operational dashboard where FTL sees new positions, company profiles, supporting evidence, opportunity priority, research, solution hypotheses, matched assets, contact routes, draft status and interaction history as one traceable record.

What was built

A scheduled agentic pipeline that discovers DACH roles, normalizes and filters positions, researches each relevant company, designs an evidence-bound response, matches approved FTL work, identifies a legitimate contact route and prepares the final outreach for exact human approval, sending and tracking.

Dual lens

Experience / System

The outcome and the infrastructure remain part of the same project truth.

The FTL Opportunity Intelligence control room showing operational metrics, recent pipeline runs and audited decisions

What people see and feel

A market in motion

The dashboard turns scattered public activity into a living field of companies, positions and capability signals that FTL can inspect each morning.

The reason is visible

Every opportunity explains which position created the signal, what the company appears to be building and why a specific FTL capability could matter.

One decision surface

Research, solution direction, matched work, contact route, message state and interaction history remain connected instead of disappearing across tabs and documents.

System

Daily DACH discovery. Scheduled searches and watched public sources find new or materially changed positions, normalize them and retain their provenance.

Evidence becomes fit. The system filters relevant signals, assembles cited company intelligence and connects the company’s direction to reviewed FTL offers, claims and project assets.

Human authority at the boundary. Agents prepare the research and message, but FTL selects the route and approves the exact version before the system sends and records the interaction.

Human + technical orchestration

Direction remains visible at every stage.

Human responsibility

Define the target region, strategic goals and capability vocabulary
Approve FTL knowledge releases, claims and externally usable assets
Review qualification, evidence and the proposed contribution
Select the legitimate contact route and approve the exact message
Own the relationship and interpret replies

Technical responsibility

Run daily discovery and monitor known job sources
Normalize, deduplicate and filter positions against relevant capability signals
Build cited company dossiers and decomposable opportunity scores
Design solution hypotheses and match zero to two eligible FTL assets
Draft the evidence-bound message, send it after approval and record the interaction

Iteration & validation

Every company fact and position claim remains linked to a public source. Filters and scores are deterministic, public research is separated from private FTL knowledge, confidential or stale assets are excluded before matching, contact routes are never guessed, and suppression and approval are checked again before sending. A model cannot qualify, approve or send on its own.

Output & deployment

The system runs inside FTL’s private operational environment. Discovery starts every morning on a Europe/Berlin schedule; the team reviews the qualified paths, authorizes the exact message, and the platform sends and records the approved interaction.

Reusable capability

The reusable outcome is not a lead list or generic CRM template. It is an architecture for turning an organization’s own signals, knowledge, assets, rules and human decisions into a governed operating system.

Learning

Agentic workflows become valuable when they begin with a real operating need and remain observable enough for a team to understand, challenge, govern and extend.

Inside the operating system

A market signal becomes a meaningful conversation.

Seven bounded stages connect public evidence, private FTL knowledge and human authority. Every handoff, decision and correction remains visible.

01
Observe
Input
Reviewed search definitions, public job sources, region and capability criteria
Output
Immutable new or materially changed job records
Human
Defines the target market and capability vocabulary
System
Searches DACH each morning, watches known endpoints, and normalizes and deduplicates positions
Decision
Is the source public, safe and new or materially changed?
Correction
Reject unsafe sources, retry failed retrievals and preserve prior valid state
Reusable · Discovery and source ingestion
02
Qualify
Input
Normalized position snapshots and exact evidence
Output
Ranked opportunity signals
Human
Defines relevance and commercial thresholds and may override a decision with a recorded reason
System
Extracts source-bound capability evidence, filters matching positions and computes decomposable scores
Decision
Does the position reveal a relevant capability need?
Correction
Retract false positives or re-run classification under a newer policy
Reusable · Signal ontology and scoring
03
Research
Input
Qualified company, position and supporting evidence
Output
Cited company dossier
Human
Authorizes research and reviews strategic context
System
Reads official company pages and bounded public sources, registers citations and assembles structured claims
Decision
Is every relevant fact current and source-bound?
Correction
Preserve unknowns instead of guessing and refresh stale research
Reusable · Company intelligence
04
Design the Fit
Input
Company dossier, position goal and active FTL knowledge release
Output
Solution hypothesis and zero to two matched assets
Human
Curates offers, claims and assets and approves the proposed contribution
System
Connects the company’s direction to an FTL response and excludes confidential, stale or unapproved material before matching
Decision
Is the proposed contribution credible, useful and supported?
Correction
Revise the solution or return a valid zero-asset result
Reusable · Knowledge, solution and asset matching
05
Find the Route
Input
Approved solution and registered company sources
Output
Buyer-role hypothesis and eligible public contact route
Human
Reviews outreach eligibility and legal status and selects the exact route
System
Infers relevant role categories and extracts only explicitly published routes
Decision
Is the route legitimate, current, unsuppressed and permitted?
Correction
Never guess an address; a complete no-route result remains valid
Reusable · Contact intelligence
06
Compose
Input
Immutable opportunity packet containing approved evidence, solution, route, claims and assets
Output
Structured subject, body and short-message units with exact bindings
Human
Sets the communication goal and tone and edits, accepts or rejects the result
System
Drafts from the position, company information, company direction and approved FTL material, then checks factual and tonal consistency
Decision
Is every claim supported and is the message specific to this opportunity?
Correction
Return individual content units to drafting or evidence review
Reusable · Drafting and factual review
07
Send & Learn
Input
Human-approved exact message version and selected route
Output
Recorded outbound interaction and follow-up state
Human
Authorizes the exact recipient, route and rendered message
System
Rechecks suppression and version hashes, sends the approved message and records the interaction
Decision
Is this still the exact approved and eligible version?
Correction
Block stale or changed inputs and preserve the complete audit trail
Reusable · Approval, outbound delivery and learning loop

Evidence & limits

Claims carry their context.

Observed

FTL uses the internal workspace to connect scheduled discovery, company profiling, source-bound research, solution and asset matching, contact routing, exact-version approval, dispatch and interaction history in one observable pipeline.

What becomes possible next

The same pattern can be transferred into client-specific research, partner scouting, business development, innovation, content or internal knowledge workflows—adapted to their sources, controls and teams.