PH

Demos

Solution demos

Recorded walkthroughs of solutions I've designed and built, set up the way I'd present them to a customer: the business problem, the solution, and the design decisions behind it.

Featured demo · Sep 2026

AeroWave Claims Concierge

Built on the Salesforce Platform

AgentforceData 360Service CloudFlowSlackEnhanced Chat

Prefer YouTube? Watch it on YouTube

The business problem

Before a flight-delay claim gets paid, someone has to confirm who the customer is, confirm the flight was actually disrupted, and check for fraud. When the customer's loyalty and booking records live in separate systems, that work is slow and manual, and customers wait for an answer.

What you'll see

  1. 01An Agentforce claims agent verifies the customer, checks flight status, and pays clean claims in seconds.
  2. 02Data 360 identity resolution unifies loyalty and booking records from 5 source systems. Weak matches go to a Service Cloud case with an agent-written reviewer brief.
  3. 03A Slack review loop where reviewers pull in the fraud team, approve or deny the claim, and the decision is written back to the case and the customer's status reply.

How it fits together

ChannelCustomer starts a claim in Enhanced Chat
AgentAgentforce claims agent
VerifyData 360 identity resolution across 5 sources
CheckFlight status lookup
DecideFlow scoring & payout logic
Clean claimPaid in seconds
Weak match or riskService Cloud case + agent-written reviewer brief
Human reviewSlack: reviewer + fraud team approve or deny
Close the loopDecision written to the case and the customer's status

Design decisions

Why it's built this way.

The agent never reveals the rules

Scoring and payout logic live in Flow, not in the agent's prompt, so the agent never exposes scores or thresholds that a bad actor could learn to game.

One customer, five systems

Identity resolution in Data 360 gives the agent a single view of the customer. When the match is weak, the agent hands off to a human instead of guessing.

Review where people already work

Reviewers approve or deny in Slack, with the fraud team a mention away. The decision flows back to the case and the customer automatically, so no one has to re-key it.

What I'd ask a real customer first

The demo shows one design. In a real engagement, discovery decides the thresholds, data sources, and handoffs.

  • How many claims come in, how long does a decision take today, and where do claims stall?
  • Who has payout authority, and above what amount must a human approve?
  • Which systems hold booking and loyalty data, and how reliably do customer records match across them?
  • Which fraud signals do you already trust, and who owns the threshold decisions?
  • What policy or regulatory rules govern refunds and compensation in your markets?
  • Which channels do customers use today, and which one should the agent start in?

Get in touch

Want a walkthrough for your use case?

I'm happy to walk through the build, the tradeoffs, or how I'd adapt it.