A huge part of our value at Ascend is being able to identify optimised routing strategies and favourable fare conditions that most travellers — and even most agencies — would never find. An agent would receive a request on WhatsApp, create a ticket, and spend up to 60 minutes across various portals and OTAs trying different routing combinations, dates, cabin mixes, and points programmes for a single request. That's fine at low volumes — but as we started to scale, it became a big bottleneck for how many clients we could serve and how thoroughly we could search.
Guru automates this. A natural language request comes in from a client on WhatsApp, gets routed through to Asana where a ticket is created, and is then sent through to Guru. Ranked flight options come out. Instead of one person searching one route at a time, Guru explores 1,000+ routing permutations across multiple providers simultaneously — variations that would take an agent hours to check manually.
The results speak for themselves: up to 70% savings vs retail business class fares, delivered in minutes rather than an hour.
How it works
Requests start as plain language, coming straight from WhatsApp. Guru parses them into structured search parameters — origin, destination, cabin, dates, airline preferences — and fans out across providers.
Most online travel agencies search direct routes on a single provider. Guru takes a different approach: it explores far more routing permutations across multiple airports and providers, looking for pricing inefficiencies that only appear when you search creatively. Think of it like TrainLine's split ticketing for UK trains — where buying two separate tickets for the same journey can be significantly cheaper than one through ticket. The same principle applies to flights, but across a much larger search space.
Results
Unlike OTAs, the "Best" ranking is personalised to each client — factoring in airline loyalty programmes, cabin preferences, and route preferences. A MileagePlus Gold member who values nonstop United flights gets different rankings than someone who just wants the lowest price.
The pipeline
The full workflow runs end-to-end without manual intervention.
A request comes in on WhatsApp, a ticket is created on Asana, Guru searches 1,000+ routing variations, ranks the best options against that client's preferences, and posts the results to Slack. The concierge reviews and sends the quote back to the client. What used to take up to an hour of searching now takes about 4 minutes of compute time and a quick human review.
Under the hood
Much of this is proprietary, so we can't share too much. At a high level, Guru combines direct API calls with robust agentic AI-powered scraping. Special shout out to Browserbase — we run up to 90 concurrent virtual browser sessions overlaid with agentic AI, which makes scraping as reliable as a direct API call.
Where we are now
Guru is in production, handling live client requests. It's already replaced the manual search workflow for the majority of bookings.
On the roadmap: airline and alliance filtering (so clients can restrict searches to their preferred programmes), passport and visa eligibility checks (automatically excluding routings that require transit visas a client doesn't hold), and deeper integration with client preference profiles so rankings improve over time.
The goal is straightforward: every search should be faster, cheaper, and more thorough when Guru does it.