Public data tells you more than most sales teams realize. Job postings, patents, SEC filings, and conference talks reveal when a company is ready for your solution.
Open-source intelligence (OSINT) is the collection and analysis of publicly available information to produce actionable insights. In cybersecurity, OSINT is used to identify threats. In B2B sales, we use the same discipline to identify opportunities.
The sources are all public: job postings on LinkedIn and company career pages, patent filings in USPTO and EPO databases, SEC/regulatory filings, conference presentations and proceedings, press releases, trade publication coverage, and open government procurement records.
What makes OSINT powerful in B2B is not any single source, but the combination. A job posting for an "embedded systems engineer with FPGA experience" tells you something. That same posting, combined with a recent patent filing for a new sensor architecture and a conference talk about edge computing challenges, tells you exactly what product that company is building and where they need help.
Individual data points are noise. Combined signals are intelligence. The Model T pipeline uses a structured approach to stack signals from multiple sources into a timing opportunity assessment.
Here is a concrete example. A company posts 3 embedded engineering positions in 60 days (hiring signal). Their CTO presents at a conference about challenges migrating from ASIC to FPGA (technology signal). A competitor announces a product in the same space with a 6-month head start (competitive signal). Their latest annual report shows R&D spending up 18% year-over-year (budget signal). These four signals together tell you: this company is building something new, they are understaffed, they face a competitive threat, and they have budget. That is a timing opportunity.
The signal stack is scored on three dimensions: urgency (how time-sensitive is the opportunity), fit (does it match Promwad's competencies in hardware, embedded, and product development), and accessibility (can we reach a decision-maker within 2 weeks). Only opportunities that score above threshold on all three dimensions proceed to concept creation.
The research phase follows a structured sequence. First, the business analyst creates a company profile: industry position, revenue, product lines, key competitors, technology stack, and organizational structure. This takes approximately 24 hours and produces a 3-5 page document.
Next comes stakeholder mapping. The analyst identifies decision-makers (VP Engineering, CTO, Head of Product), their professional backgrounds, communication preferences, and the internal dynamics that influence purchasing decisions. This psychological profiling uses public data: LinkedIn career history, published interviews, conference presentations, and patent authorship.
The technology gap analysis compares the company's current capabilities against market requirements. Where is the gap between what they have and what they need? The competitive landscape analysis maps this gap against what competitors offer. The opportunity scoring combines all inputs into a prioritized list of product concepts worth building.
The volume research problem in B2B is real. A single company profile requires reading 20-50 web pages, cross-referencing 5-10 databases, and synthesizing information from multiple languages. Doing this manually for 5-7 target accounts would take a single analyst 2-3 weeks. With AI assistance, the same analyst covers the same ground in 3-5 days.
AI handles the volume tasks: scraping public job postings, summarizing patent filings, extracting financial data from annual reports, and drafting initial company profiles. The analyst focuses on high-judgment tasks: interpreting signals, identifying timing opportunities, and making Go/No-Go decisions on which concepts to pursue.
The Model T pipeline uses 8 specialized AI agent roles, each with a defined scope and quality checklist. The business analyst role handles steps 4-6 (company profiling, insight generation, technical review coordination). The product manager role handles steps 7-12 (insight selection, deep analysis, concept creation). Every AI-generated output is reviewed by a human before it moves to the next stage.
Model T uses public data only. No private databases. No social engineering. No pretexting. No scraping of password-protected content. No purchasing of leaked data. The boundary is clear: if the information is not available to anyone with a web browser, we do not use it.
This is not just an ethical stance; it is a practical one. Public data is defensible. When a client asks "How did you know about our FPGA migration project?", the answer is "Your CTO presented about it at Embedded World 2025." That answer builds trust. "We bought a list from a data broker" does not.
The stakeholder profiles we create contain only information that the individuals themselves have made public: LinkedIn profiles, conference talks, published papers, patent filings. We do not monitor personal social media, private communications, or non-professional activities. The goal is to understand professional priorities, not personal lives.
We use a combination of AI-assisted research tools, patent databases (USPTO, EPO, WIPO), financial databases (SEC EDGAR, Companies House), LinkedIn for professional profiling, and industry-specific sources like trade publications and conference proceedings. No proprietary or paid intelligence databases are used for client-facing research.
Private companies with minimal web presence are harder to profile but not impossible. Patent filings, regulatory submissions, trade show attendance, and hiring patterns still provide signals. If the combined signals are insufficient to build a credible concept, the company is filtered out at the scoring stage rather than receiving a low-quality deliverable.
Yes. GDPR governs the processing of personal data, but it explicitly allows processing of data that individuals have made publicly available for legitimate business interests. Our stakeholder profiles contain only professionally published information. We do not build profiles from private or scraped data, and we do not engage in automated individual decision-making.