By Keshavan Thiagaraja
Planning a family holiday can take hours before a single booking is made. Flights, rooms, meals and entry requirements must fit together. A cheap flight may have an exhausting connection. A suitable hotel may not permit three people in one room. A vegetarian meal may still contain meat broth.
Instinct AI brings a practical question into focus: how much of this work can we delegate, and what must remain under our control? For CEOs and CTOs, the same question applies to purchasing, customer service and other enterprise workflows.
What Instinct is offering
Founded in 2025 by Noah Shinn, who previously worked at Sierra, Instinct is developing a personal AI assistant that accepts instructions through text or calls. Its stated capabilities include planning trips, buying groceries, booking tickets and cancelling subscriptions.
The company says its assistant uses a phone and computer to complete tasks. In its September funding announcement, it also described a concierge service that can call businesses on a user's behalf. These are company descriptions of capability, rather than independent guarantees of performance.
On 28 September 2026, Instinct announced US$1 billion in funding at a US$10 billion valuation. Reporting put its team at 14 employees. That is a striking combination, but valuation measures investor expectations. It does not establish profitability, service quality or the economics of each completed task.
A family holiday makes the challenge concrete
Consider a holiday I would like to plan with my wife and our 17-year-old daughter. We would leave Colombo on 20 December 2026 and must arrive back on 28 December. We would like to visit Beijing and Shanghai, including the Great Wall, at a relaxed pace.
We prefer comfortable three-star or four-star hotels near public transport, with a family room for three. We would also like comparable five-star prices. I am 100% vegetarian. I do not eat eggs, meat, fish, seafood or meat-based broth.
This is an illustrative task. I have not tested this holiday through Instinct.
A useful assistant should first ask for our total budget and essential missing information. It should assess whether two cities are practical once international flights, transfers, travel between cities and rest are included.
Flight comparisons should cover the return fare for all three people, baggage, connections and total journey time. Hotel comparisons should show taxes, breakfast, cancellation terms and actual bedding arrangements, including whether our daughter is charged as an adult.
Entry and transit requirements need separate checks for each passport using official sources relevant to our travel dates. The assistant should flag uncertainty rather than turn an incomplete check into a confident assurance.
The itinerary should consider winter conditions and advance reservations. Its budget should include flights, accommodation, transport, food, attractions, insurance and any visa costs. Confirmed prices and estimates should be clearly distinguished.
Compare the work and the responsibility
A recommendation assistant suggests a hotel. An action-taking assistant must confirm occupancy, obtain approval and establish whether the reservation succeeded.
A recommendation assistant lists flights. An action-taking assistant must preserve the return-date requirement through checkout and verify the supplier's confirmation.
A recommendation assistant produces a plan. An action-taking assistant must track unfinished steps, explain failures and support recovery.
This comparison matters because the responsibility grows with the authority granted. Better language alone does not make the execution dependable.
Where the value and risks meet
The immediate opportunity is time saved. One conversation could reduce repeated searches, comparisons and calls. Personalisation could improve the result by carrying dietary requirements, room preferences and travel pace through every decision.
But a misunderstood date or unsuitable room can cost money. A reservation attempt is not a confirmed booking. Users need supplier confirmation and a clear record of what was authorised.
Privacy also becomes a design decision. Useful assistance may require email, location, travel documents or payment access. Users should understand what is accessible, for how long, and how to revoke permission.
Commercial incentives deserve equal attention. Skift reported that Shinn discussed merchant commissions as a possible business model. That does not establish how every current recommendation is ranked. It does show why users should be told when a commercial relationship could influence a choice.
What CEOs and CTOs should do
CEOs should select a narrow workflow with a measurable cost today. Compare staff time saved with the full cost of software, supervision, exceptions and rework. Judge a pilot by correctly completed tasks, not the number of convincing responses.
CTOs should define permissions before connecting systems. Start with limited access. Require explicit approval for purchases and other consequential commitments. Approval should identify the supplier, amount and terms the user is accepting.
Keep a record of instructions, approvals, tool actions and supplier responses. Distinguish prepared, approved, submitted and confirmed states. Prevent retries from creating duplicate transactions. Provide a human route when a task fails or a requirement remains uncertain.
Together, leaders should test awkward cases before expanding access: conflicting dates, changed prices, ambiguous room occupancy, unavailable services and withdrawn permission. These are ordinary operating conditions, not edge cases to postpone.
The holiday example is a useful model for enterprise adoption. The assistant must carry a person's requirements across several systems without losing the conditions that made the task worthwhile.
The approval that matters
My instruction for this trip would be clear: “Present the itinerary and booking options for my review. Do not book, pay or make any financial commitment without my explicit approval.”
Instinct's appeal is understandable. Many people would welcome help with the tasks that consume their evenings and weekends. Its lasting value will depend on reliable execution, understandable choices and respect for the authority each user grants.
Which workflow would you trust an AI assistant to complete, and what evidence would you require before allowing it to act?
Think mathematically. Engineer intelligently. Lead strategically. Build responsibly. Scale globally. Create real business value through artificial intelligence.
#ArtificialIntelligence #AgenticAI #InstinctAI #AIStrategy #EnterpriseAI #ResponsibleAI
Originally published on LinkedIn by Keshavan Thiagaraja. Read the original on LinkedIn.
