Sightline

The context

I lead applied AI technology for Sightline, PwC’s global tax platform. The work spans architecture, engineering, release governance, production operations, and teams distributed across the network.

Sightline serves practitioners and clients through a controlled-access platform. Its quality bar is straightforward: reliable, secure, usable, supportable, and fit for a tax context.

The leadership challenge

Tax answers are only useful if they are right. When Worldwide Tax Summaries returned incorrect jurisdictional dates, that was not a polish issue. We needed to pause the feature, fix the source and retrieval path, and prove it had earned its way back.

What I led

I led delivery for Sightline AI Assistant, bringing together tax Q&A, drafting and summaries, and document search across multiple production releases.

I helped lead the Worldwide Tax Summaries recovery. We paused the feature, corrected ingestion and territory mapping, expanded validation, and restored it only after the work had been checked across the jurisdictions it served.

I also co-led practical AI enablement through pilots, live workshops, and delivery improvements that pair AI-assisted work with deterministic checks. The point was not more AI activity. It was dependable software and useful evidence when it mattered.

The result and lesson

The feature returned to global use after its source and retrieval path were corrected and validation was expanded.

On a consequential AI system, the payoff is not just a better answer. It is a release path that can stop, correct, and prove when the answer is ready.

Technology and operating environment

• Controlled-access tax platform
• Release governance
• Retrieval and ingestion mapping
• Validation-led restore
• Agentic loops with deterministic checks