Face2Face.

An AI-powered sales assistant that transcribes meeting audio and analyzes it against quality and compliance rules, helping promoters close more deals while avoiding risky behavior.

AI coaching for high-volume field sales.

Face2Face Design & Communication is a Nordic sales and field-marketing company that equips promoters with tools and training to run high‑volume, compliant door‑to‑door and in‑store campaigns. The company operates across multiple markets and campaigns, with a strong focus on sales quality, legal compliance, and data-driven coaching.

The core business objective behind this initiative was to augment sales performance and compliance through AI‑assisted speech analytics and real‑time coaching at scale.

What the client needed.

The client set a goal to build an AI‑powered sales assistant that transcribes meeting audio and analyzes it against quality and compliance rules to help promoters close more deals while avoiding risky behavior. They needed:

Reliable transcription that works well on iPads and, at times, phones. Strong non‑English support was essential.

A rules framework able to handle both standard and campaign‑specific policies, with a path toward real‑time objection handling and guidance.

GDPR‑compliant processing with tight control over data retention and vendor DPAs, and awareness of the EU AI Act constraints on employee evaluation.

Hard problems in capture, architecture and legacy code.

Getting accurate, multilingual transcription on target devices meant solving three problems first.

Diarization & capture flow constraints: speaker diarization limitations in available models and browser constraints impacting capture flows, particularly on Safari and iOS.

Choosing the right transcription architecture: navigating trade‑offs between on‑device and streaming transcription, and picking a provider that balances cost, accuracy, multi‑language quality, and compliance.

Modernizing legacy PHP with Laravel APIs: evolving a legacy PHP backend toward modern APIs in Laravel, and planning for scalable rule management and analytics.

What we delivered.

Generic Soft was engaged to prototype, evaluate, and implement the AI transcription and rule‑based analysis components, integrate them with the client’s workflows, and chart a path for real‑time coaching. We collaborated on provider selection, legal readiness, front‑end experiments, and back‑end evolution.

Transcription provider selection and integration plan, choosing DeepGram after testing and GDPR review, with DPAs for both DeepGram and OpenAI.

Audio capture and streaming experiments, including a Flask prototype for chunked streaming to DeepGram and browser‑security workarounds.

An analysis output flow that records applied rules in generated analysis JSON for traceability.

A working demo application with minimal UI and a front‑end rule selection mechanism, designed to evolve toward database‑driven rules at the campaign level.

Robustness improvements: enhanced error logging and retry logic for model interactions.

Compliance grounding: confirmation of GDPR approach with CFO and guidance on data retention and EU AI Act implications.

A compliant path to accurate, multilingual transcription.

The team established a compliant, practical path to accurate, multilingual transcription on target devices, unblocking the MVP and creating a blueprint for real‑time objection detection and rebuttal suggestions. Operational reliability improved through better logging and retries, and the rules framework now has a clear migration path from hard‑coded to campaign‑managed configurations.

These steps set the stage for higher sales quality at scale and safer compliance across campaigns.

Built with

More case studies

Ready for results like Face2Face's? Discover if we're a good fit for you.