We tested whether artificial intelligence could eliminate that overhead entirely and built a working Travel AI assistant in four hours. This article documents what we built, how it works, and why the approach extends beyond travel coordination.
The intake bottleneck in corporate travel management
The inefficiencies we encountered are not unusual. They reflect a structural condition in corporate travel management that persists across organisations of all sizes, one that significant software investment has not adequately resolved, because most tools were built to process travel requests, not to gather the information needed to initiate them. Industry data confirms this is a widespread pain point:
The problem is also growing. According to Deloitte's 2025 Corporate Travel Buyer Survey, three in four travel managers report expanding budgets, meaning coordination teams are being asked to handle more volume with the same administrative infrastructure. The gap between workload and capacity is widening, and the bottleneck consistently sits at the same place: intake.
The hidden cost of unstructured information collection
When organisations discuss AI in the enterprise, the conversation typically focuses on decision support, analytics, or automation of rule-based tasks. Rarely does it address what is arguably the most expensive hidden cost in knowledge work: the overhead of collecting structured information from people.
In our case, the travel coordination team was not slow because it lacked expertise. It was slow because each request meant chasing employees across multiple channels to assemble facts that should have been captured in a single interaction. The internal numbers made that cost concrete:
- 70% of the travel team's time was spent collecting information from employees, not coordinating travel.
- 3 days were required to complete a single visa request due to back-and-forth message exchanges.
This is a pattern that exists in virtually every professional services organisation: HR onboarding, IT helpdesk, procurement approvals, and legal intake.
The expert's work takes an hour. The information collection takes a week.
Why existing tools don't solve the information collection problem
Three categories of tools are commonly used to solve travel coordination inefficiencies. Each addresses a real symptom but does not reach the root cause. They were designed to process structured information, not collect it from people who have not yet provided it.
| Approach | What it promises | Where it fails |
|---|---|---|
| Portal / form systems (SAP Concur, ServiceNow) | Structured data capture | Requires the employee to already know what's needed. Rigid. No guidance. |
| Rule-based chatbots | Scalable self-service | Cannot handle context-dependent processes with 50+ data points and variable logic. |
| Knowledge bases (Confluence, SharePoint) | Reduce inbound questions | Passive repositories. Employees bypass them entirely and ask the team directly. |
Under the hood: what powers the Travel AI Assistant
The Travel AI Assistant we developed in four hours was not a booking system. It was a proof-of-concept for a different category of enterprise tool: conversational process automation.
The system uses three capabilities that, combined, address the information collection problem at its root:
1. RAG for institutional knowledge
Using Retrieval-Augmented Generation (RAG), the assistant indexes internal policy documents and answers questions about them directly with source attribution, instead of expecting employees to find the right Confluence page. Per diem rates, approval thresholds, and required documentation are instantly accessible within the natural flow of conversation. Employees stop bypassing documentation not because their behaviour changed, but because the friction of finding it was removed.
2. Context-aware information collection
Visa requirements vary by citizenship, destination, and travel purpose. A conversational AI can adapt to this complexity that traditional forms struggle with. The assistant dynamically adjusts its requests based on what it already knows, prefilling available data from employee profiles and surfacing only what's missing.
3. Web search with source transparency
The AI Travel Assistant doesn’t rely on outdated internal data. This addresses the compliance risk that arises when teams use generative AI tools like ChatGPT for research without systematic verification.
The result: A single 5-minute conversation replaces 10+ message exchanges over three days. The travel team receives a complete submission, not a fragment that requires follow-up. Policy questions that previously reached human inboxes are answered instantly, at any hour.
What a 4-hour prototype delivered
We are deliberate about not overstating the scope of what was built. This was a proof-of-concept, not a production deployment, but its value lies precisely in demonstrating feasibility at low cost. Across four dimensions, the results were consistent:
The logic for implementing this class of solution is straightforward:
- Immediate ROI: 30–50% reduction in time spent on information collection by expert staff. Measurable against salary costs from day one.
- Quality improvement: Complete, validated submissions versus the typical 60% completion rate with manual collection. Fewer errors, fewer correction cycles.
- Scalability: The system handles concurrent requests without linear growth in headcount. A team of three continues to serve two thousand employees as the organisation expands.
- Employee experience: Friction reduction is not a soft benefit. Multi-day back-and-forth on administrative requests has a measurable impact on skilled employees' willingness to pursue international opportunities.
Beyond travel: other enterprise workflows with similar bottlenecks
Travel coordination was our test case. The pattern it validates applies far more broadly. Any enterprise workflow that currently follows this sequence: employee submits incomplete request → expert chases missing information → information arrives in fragments → expert assembles, and acts is a candidate for this approach. The expert's judgment is not being replaced. The 70% of their time is spent on pre-work that doesn't require expert judgment.
HR onboarding. IT helpdesk triage. Procurement intake. Legal matter opening. Compliance documentation. In each case, the bottleneck is the same, and the solution architecture is identical: a conversational layer that structures the chaos before it reaches the human who knows what to do with it.
The components, large language models, RAG pipelines, and web search integration are available, mature, and composable. The prototype we built in four hours reflects not just the capability of current AI tooling, but the clarity that comes from understanding exactly which part of a workflow is worth automating.
Conclusion
What this prototype demonstrated is that most of the travel team's time spent collecting information before any real coordination begins is automatable today, with available tooling, in a realistic timeframe.
The gap between that insight and a working implementation is not a technology problem. It is a process design problem: understanding which information needs to be collected, in what order, and in what form. That is the work worth doing and where the returns are most significant.
Modern AI tooling has reached an inflexion point where significant workflow automation is achievable within proof-of-concept timeframes. The question is no longer whether it's technically possible. It's which processes your organisation will automate first.
Related Insights
The breadth of knowledge and understanding that ELEKS has within its walls allows us to leverage that expertise to make superior deliverables for our customers. When you work with ELEKS, you are working with the top 1% of the aptitude and engineering excellence of the whole country.
Right from the start, we really liked ELEKS’ commitment and engagement. They came to us with their best people to try to understand our context, our business idea, and developed the first prototype with us. They were very professional and very customer oriented. I think, without ELEKS it probably would not have been possible to have such a successful product in such a short period of time.
ELEKS has been involved in the development of a number of our consumer-facing websites and mobile applications that allow our customers to easily track their shipments, get the information they need as well as stay in touch with us. We’ve appreciated the level of ELEKS’ expertise, responsiveness and attention to details.
