An entire investment team building its own research agents in one day

Client

Date

Corporate venture arm of a $10B+ global electronics manufacturer

08/2026

The challenge

The corporate venture arm of a $10B+ global electronics manufacturer invests in early-stage technology companies. Roughly forty people carry a repetitive research load: a market map before entering a space, a dossier before every founder call, a diligence pass on every candidate, a weekly portfolio briefing. All of it assembled by hand.

The solution

FifthRow put the platform in the hands of the firm's own people rather than delivering research to them. The team built its own agent systems for the jobs it repeats, then ran them against live deals, live portfolio work and a live weekly publication.

11,805

analyst-hours equivalent, 8 FTEs

17

research systems the team owns

Standing research systems17, built by 12 of the firm's people

Weekly briefingStanding weekly portfolio and media pass

Analyst-hours equivalent11,805 hours saved; equivalent to 8 FTEs

01 — ONE-DAY BUILD WORKSHOP

The engagement opened with a one-day workshop, the firm's own people and the FifthRow team in the same room, building systems rather than being handed research. By the end of the day thirty-one of them had run an agent and twelve had built one.

Core Activities

Ran the workshop on site over one day, the FifthRow team alongside investment, portfolio, comms and operations
Taught the governing distinction: an agent does a job, a system sequences agents, an app covers a whole process
Had each function build for its own job, not a common template: weekly news, portfolio summaries, deal summaries in a target sector, candidate evaluation, a red-team agent against a deal, an anti-portfolio tracker, a proposal builder, a grant finder
Left the systems in the team's own accounts, editable by whoever built them
Iterated investors' own automations against live deals later in the year

The Results

Seventeen working systems, built by twelve different people across four functions, all of them named after a job the firm actually does rather than a feature of the platform.

of the firm's own people, day one

02 — SCHEDULED BRIEFING PRODUCTION

The firm publishes a weekly internal briefing covering the prior week's portfolio milestones, investment announcements, media coverage and speaking engagements. It reaches the investment team, the engagement team and the external PR teams, and its introduction is copied into emails to the parent group's chief executive.

Core Activities

Took the author's written spec section by section, including the priority order behind the lead story
Swept portfolio milestones, investment announcements, media coverage and thought leadership for the week just gone
Set it to the publication's own weekly cadence, so the draft exists before the author starts
Replaced a manual sweep of two internal boards, a PR morning briefing and a watch list
Kept human verification deliberately, since the publication's accuracy is under direct scrutiny

The Results

A drafted weekly briefing built from a source sweep no person had time to do by hand, with the author's judgement and verification applied on top rather than underneath.

3,545

analyst-hours equivalent, ~2 FTEs

03 — DEAL SOURCING & MARKET MAPPING

Before a corporate investor enters a space it needs a map of it: who is building, what is funded, where the incumbents are weak. The firm used the platform to produce those maps on the technology areas its parent group cares about.

Core Activities

Ran market analysis on the parent group's closest technology areas, one map from 1,275 sources
Framed opportunity reports as competitive questions, not topics: what it would take to displace the established players in a category
Ran scientific-news-driven sourcing for companies under five years old in a target technology
Returned currently available grant funding by sector and geography
Kept passed deals visible and reviewable in an anti-portfolio tracker

The Results

Standing, repeatable market maps on the technology areas the parent group invests behind, produced from a named market and region rather than from a research brief.

1,275

sources behind one market map

04 — DILIGENCE & DEAL SCRUTINY

The heaviest research load in venture is the work done between a first meeting and an investment decision. The firm used the platform on the parts of that load that repeat on every deal.

Core Activities

Got a full read on a candidate in one pass: company, founders, patents, competitors, exit paths
Ran technology and intellectual property analysis on individual candidates
Ran pre-call briefings on demand, everything available on a company as one structured output
Argued the case against a candidate with a red-team agent, used on live deals
Applied the same evaluation criteria to every company entering the pipeline

The Results

A consistent pre-decision research pass applied the same way to every candidate, and a documented argument against each one alongside the argument for it.

1,499

analyst-hours equivalent, one deep-dive

What it proved

The engagement opened with a one-day workshop rather than a delivery. Thirty-one of the firm's people ran 313 agent runs in that day, and twelve built seventeen systems named after jobs the firm actually does. In the following months those systems ran against live work: the weekly briefing saved 3,545 analyst-hours, market maps covered the parent group's technology areas, one from 1,275 sources, and company deep dives ran as a single pass, one saving 1,499 analyst-hours. In total, 485 runs against 15,558 verified sources and 11,805 analyst-hours saved, the equivalent of eight full-time analysts over the period.

The goal was to put the firm's own people in a position to build and run the research they repeat, rather than to have it delivered to them. Four jobs with a fixed shape and a weekly rhythm now run as standing systems: the weekly briefing, the pre-call dossier, the candidate pass and the red-team argument. Each is owned by the person who uses it and runs against live deals. The one-off questions an investor asks to understand an unfamiliar value chain stay hand-run, which is where the boundary of the exercise sits.

Results

Metric	Benchmark	Result	Improvement
Standing research systems	0. Maps and dossiers by hand	17, built in-house	Nothing shared to seventeen owned systems
People able to build agentic workflows	Only a few technically advanced individuals	12 built one, 31 ran one	Whole firm exposed in a day
Weekly briefing assembly	Two boards, a PR briefing, a watch list	3,545 analyst-hours saved	Gathering automated, verification retained
Analyst-hours saved	Investors and comms staff, own desk research	11,805 hours, 485 runs	Equivalent of adding 8 full-time analysts over the period

TEAM-OWNED SYSTEMS

Research systems sit in the accounts of the people who built them, owned and editable by the investor whose job they were built for.

STANDING BRIEFING

The weekly publication is drafted from a standing source pass, so its author spends the time on verification and judgement rather than gathering.

REPEATABLE MARKET MAPS

A technology area and a region go in and a structured map comes out, so entering a space starts from evidence rather than a research brief.

PRE-CALL DOSSIERS

Everything available on a company arrives as one structured read before a founder call, rather than being assembled by whoever takes the meeting.

CONSISTENT SCREENING

Every company entering the pipeline is assessed against the same criteria, so candidates stay comparable rather than judged by whoever met them.

COUNTER-CASE AGENT

A standing agent argues the case against a candidate, so the objection is on file alongside the case for it before a decision is made.

Recent case studies

See more proof

From one-off research requests to a standing new-business intelligence practice

Client

$500M+ global speciality chemicals manufacturer

Date

08/2026

382,406

analyst-hours equivalent, 155 FTEs

Ten full-time analysts' worth of research, without a research department

Client

€4B+ European broadcasting and media group

Date

08/2026

18,398

analyst-hours equivalent, 10 FTEs

Matter More

Let's Talk