Why we built WitAI: AI-leveraged sourcing for B2B teams

Federico and I started Wisewood in 2023 with a clear thesis: B2B sourcing was structurally inefficient in ways that AI tooling was about to make obvious. Two years in, the platform we built — WitAI — has become the operational engine behind every project we run. This is the why and the how, and a sketch of where it's heading.

The problem with sourcing intake

If you've ever asked a sourcing agent for a quote, you've watched the same pattern play out. You send a brief. Three days later you get clarifying questions. You answer them. A week later you get a quote. You ask a follow-up. Another week passes. Two weeks in, you have a quote you can act on. For an industry whose entire value proposition is operational efficiency, the intake step was — and at most agencies still is — the slowest, most error-prone, most context-losing part of the workflow.

The reasons are structural. A senior project manager has to read every brief manually, decide which factories to shortlist, run the first round of price negotiation, draft the quote, and translate any technical ambiguity in the brief into a question to the buyer. They're doing this for a dozen briefs at once, and each one is unique. The bottleneck is human attention, not factory capacity. AI tooling targets exactly that bottleneck.

What we built

AI-leveraged intake

WitAI parses incoming briefs — text, sketches, reference photos, even a Pinterest mood board — and extracts the structured fields that drive the rest of the workflow: product category, target volume, budget range, certification needs, timeline. It identifies missing information and either asks the buyer or flags it for our team to ask. The whole step that used to take three days of back-and-forth happens in minutes for the buyer, with our team validating outputs rather than typing them.

Factory matching

Once the brief is structured, the platform matches it against our supplier database — 2,000+ verified factories across Asia, Europe, and Africa, indexed by category capability, capacity tier, certification profile, lead-time history, quality scores, and price-point bracket. The output is a shortlist of candidates with the rationale for each match. A senior project manager still does the final selection — AI proposes, humans choose — but they're choosing from a curated three or five, not searching from scratch.

Workflow management

Every step from sample order through production through QC through shipment runs through WitAI. Buyers see project progress on a single dashboard. Our team sees the same dashboard. QC reports, factory progress photos, freight tracking — all in one timeline. The platform is the operational record of the project, not a sales surface bolted on top of an opaque agency workflow.

The catalogue path

Alongside custom sourcing, the platform exposes a curated catalogue of products in our highest-velocity categories — promotional goods, merchandising, gadgets, event uniforms, technical textiles. These are products we already source at scale, where we hold the inventory commitment, and where the buyer can customise with their own branding without clearing factory MOQs. For a buyer with a tight event timeline or a marketing-driven launch, the catalogue path is faster than custom production by an order of magnitude. For a buyer who's never run a sourcing project before, it's a much lower-risk way to test the relationship before committing to a bespoke build.

What changed for buyers

On the custom path, time-to-first-quote dropped from weeks to days, and time-to-first-sample dropped from a month to one to four weeks. On the catalogue path, time-to-shipment dropped to as low as a few weeks for branded customisation on stock items. Buyers who used to wait two weeks just to know whether a project was viable can now find out in one afternoon. The friction at the start of the relationship is what was killing conversion at most sourcing agencies; we removed it.

What changed for us

Internally, the AI layer absorbed the kind of work that doesn't need senior judgement and freed our project managers to spend their time on what does — supplier negotiation, defect recovery, schedule rescue, the relationship work that defines the difference between a project that ships and a project that almost ships. Our cost-to-serve dropped, our buyer-satisfaction metrics rose, and we ended up with capacity to take on more projects without hiring linearly with revenue.

“AI didn't replace our team. It removed the work our team shouldn't have been doing, so they could do the work only humans can.”— Marco Legnini, Co-founder & CEO of Wit

Where it's heading

The next layer we're building is direct buyer-to-platform interaction for the catalogue path — buyers configuring branded customisations, getting instant pricing, and triggering production from the WitAI dashboard without an account-management hand-off. For custom production we're moving toward AI-assisted spec validation — catching the technical ambiguities in a brief before they become factory questions, before they become defects.

Longer-term, the goal is for buyers to feel that engaging Wisewood for a sourcing project should feel like opening a tool, not opening a service relationship. The relationship still matters — for non-trivial projects it always will — but the friction of getting started should be zero. That's what AI lets us build, and that's why we're building it.