See you at BaselHack 2026! (:
Teamname

Adobee Hive

Challenge description

Who we are


Adobe. We build the tools businesses use to create and manage their digital presence, and increasingly the systems that help them stay found and represented as AI enters the picture. This challenge comes from the team working on AI brand visibility and agent-facing content: helping brands stay accurately found, cited and represented as AI and agents become the ones doing the reading. We want the Basel community’s creativity on where this is heading, not a finished product.
The problem (context) – ADDED (Context)
Think about how a small hotel or a neighbourhood restaurant has shown up online over the last 25 years:
~2000s: everyone needed their own website. That was the presence.
~2010s–today: the website matters less. Discovery, booking, menus, reviews, photos and even “call now” all live inside Google Maps, TripAdvisor, Booking.com, OpenTable, Instagram and similar aggregators. The business fills in structured profiles; the platforms own the interface.
Emerging: the first interaction is increasingly an AI. A traveller asks an assistant “book me a quiet, dog-friendly hotel near Basel’s main train station (Basel SBB) under CHF 200 with late check-in” and never visits a website or a map at all. The agent reads, compares, decides and books.
If that trajectory continues, the primary surface stops being the human-facing website, or even the human-facing platform UI. The business still needs to be discoverable, accurately represented, bookable and trustworthy, but to a machine. Does that mean an MCP server instead of a homepage, a structured, verifiable “business object” that agents can query and trust, or something nobody has named yet?
The motivation
This is not sci-fi. It’s already happening, and nobody has a settled answer yet. We’re less interested in “yet another CMS” and much more interested in a credible, imaginative picture of the future plus a prototype that makes it tangible. Getting this right matters for millions of small businesses that can’t afford to lose the discovery game a second time. That’s the frame this challenge asks you to fill in.
The challenge — How might we… – Added (Problem Statement)
How might a local business (hotel, restaurant, café, shop, service provider) establish and control its presence in a web where the primary “visitor” is an AI agent rather than a human browsing pages?
Deliberately broad and open. Pick the angle that excites your team. Some directions to spark thinking. This isn’t a checklist or a spec, and inventing your own angle is encouraged and rewarded:
The agent-facing storefront. A business exposes its offering (rooms, tables, menu, hours, availability, price, policies) as something an agent can query and act on — e.g. an MCP server (see terms above) — instead of, or alongside, a website. What does “publishing” mean here? What does the owner see and edit?
Discovery & trust. In a world of agent-mediated choice, how does a business get found and believed? Verifiable claims, reputation, provenance (proving where a claim about your business actually came from), spam resistance, “why did the agent pick this one?”
The transaction. Agent negotiates and completes a booking/order end to end through whatever agent-queryable interface you propose. Confirmation, changes, cancellation, payment hand-off, edge cases.
The owner’s experience. The café owner is not a developer. How do they set all this up in minutes and stay in control of how they’re represented?
Representation & fairness. Who decides how a business is described to the AI? How does a small player avoid being flattened into a generic result or losing to whoever games the system?
The human fallback. Humans and agents will coexist for years. How does one presence serve both without maintaining two of everything?
What to deliver (by the end of the hackathon) - ADDED (must haves)
Two things, weighted together:
A vision. A clear, believable articulation of how local-business web presence looks in ~5–10 years — the concept, the actors, why it’s plausible. A short deck / narrative / demo storyboard is fine.
A prototype. Something that runs and demonstrates a real slice of that vision — e.g. an agent successfully discovering and booking a fictional Basel hotel through whatever interface you propose. It doesn’t need to be complete, but the core mechanism should genuinely work, not be faked end to end.
You don’t need an engineer on your team to pick this challenge. BaselHack team formation is designed to mix skill sets, and we’d rather see a small, honestly-scoped working slice than a large team’s unfinished one.
If you’re not sure where to start (optional) – Added (Nice-to-have - aber formatting ist super broken…)
Here’s one minimal path that satisfies both deliverables. Ignore it if you already have a stronger idea — it’s a floor, not a ceiling:
Invent a fictional Basel hotel or restaurant with a few rooms/tables, prices, hours and availability.
Expose that data through a small MCP server (or any simple API) an AI agent can call.
Wire up a basic AI agent — e.g. an LLM with function/tool calling — that reads the offering, compares it against a traveller’s request, and completes a mock booking end to end.
Build your vision narrative around what you’d add next (trust, discovery, payments, the owner’s dashboard) if you had more time.
The more original and ambitious your take beyond this floor, the better you score on innovation.
Evaluation criteria
We judge on three axes (roughly equal weight):
Vision & believability. Is the future you paint compelling, realistic and grounded — for the business owner as much as for the technology? Could this plausibly be built and adopted? Do the pieces (technical, human, business) fit together?
Innovation. How original and imaginative is the idea? Does it reframe the problem in a way we hadn’t considered?
Execution & craft. We want real craft behind the idea, whether that shows up as working code, a rigorously worked-through interaction design, or a business model pressure-tested against real constraints. A prototype where the essential mechanism actually works beats a polished mockup that’s clickable but hollow. Mock the boring edges if you must, but the heart of the idea, in whatever discipline carries it, should be genuinely realised, not just described.
Alongside the three axes we’ll also call out a “best vision no one expected” recognition, so a standout non-technical contribution has a named way to be celebrated.
Judging panel: Adobe engineers from the challenge team plus BaselHack organisers. Winners are selected at the closing presentations.
Who should join this challenge – Added (Who should join)
Great fit for full-stack and backend engineers, UX and product designers who like ambiguous problems, product- and business-minded people (especially anyone who has run, worked in, or studied a small hospitality or retail business), people curious about how AI agents and tools like MCP work, and anyone who enjoys “what does this even look like in five years” questions. You don’t need to have built an AI agent before. We’ll cover the basics at kickoff, and curiosity plus the drive to make an idea real — through code, design, or a sharp, well-argued narrative — matter more than a specific technical background.
Your work, your IP
You keep full ownership of everything you build. BaselHack repositories are public by default, and by entering this challenge you agree that Adobe may reference and showcase your submission — with credit to your team — in an internal readout at our Basel office after the event. We will not commercialise your work or claim ownership of it. If you’d prefer we not feature a specific piece, tell a mentor and we’ll respect that. Please include an OSI-standard license (e.g. MIT) in your repo so this is unambiguous.
Contact
Damian Zehnder — zehnder@adobe.com 

BaselHack