Show the system. Then show what changed.

Challenge

Inbound inquiries arrived through six forms, an intake team of one re-keyed records, contracts were assembled case by case, call-center bursts threatened platform limits, and signed cases had to reach partner law firms without paralegals re-keying data.

System

GAT designed a normalized intake model in Salesforce and Litify, buffered Five9 call-center bursts behind Platform Events so lead routing survives peak volume, built per-litigation qualification questionnaires (the largest a 4,548-line flow) with eligibility rules in custom metadata a business user can update, mapped Docrio templates to record data ahead of Adobe Sign delivery, and shipped a distribution waterfall that auto-delivers signed cases and documents to 8+ partner-firm systems with per-firm auth and resend recovery.

Business change

The intake team grew from a single coordinator to a multi-seat operation on the new workflow, and during the system's operating period monthly intake revenue rose from $10K to more than $360K. That is a client before-and-after result, not a claim that software was the only growth variable.

Challenge

A surgical-care navigation company was paying for a standalone messaging vendor while its clinical data lived in Salesforce. Leaving the vendor meant replacing real-time chat, preserving years of history, and serving Spanish-speaking patients without adding staff.

System

GAT architected a real-time, HIPAA-conscious patient–nurse chat on Platform Events and CometD inside Experience Cloud — read receipts, attachments, scheduled messages — migrated the historical message archive with our own ETL, added Gemini-powered live translation so nurses type English and Spanish-speaking patients read Spanish, and built an access-audit trail that has logged 34,860 record views.

Business change

Five years on, the platform manages 100,000+ patient messages, supports 15,314 surgeries, and serves 51 active users on an org with 166 authored Apex classes, 59 Lightning Web Components, and 2 authored Platform Event architectures. The message count includes migrated history — it measures the system of record, not chat throughput alone.

Challenge

Calls were not logged consistently, follow-ups depended on memory, and enrichment from Apollo.io and ZoomInfo sat outside the CRM. Managers could not see whether a lead was worked or merely assigned.

System

GAT connected Aircall CTI to Salesforce, logged calls against the right records, converted dispositions into follow-up tasks, brought enrichment into the lead flow, and surfaced call volume, compliance, and conversion in dashboards.

Business change

Client reporting showed 50% more leads converted after launch, with follow-up compliance above 95% in the first month. We present this as observed post-launch performance; rep execution and other commercial factors remained part of the result.

Challenge

Every new patient was typed twice — once into the EHR, once into Salesforce — and the two systems drifted. Appointment changes, new patients, and AI-answered phone calls all needed to land in the CRM without a human re-keying them.

System

GAT built EHR webhooks landing on Apex REST endpoints that publish 38-field patient Platform Events; Flows subscribe and update Person Accounts and appointments in real time, while outbound triggers push new patients back to the EHR with MRN writeback and full error logging. The practice's voice-AI phone agent was wired into the same pattern — AI-answered calls become structured Leads — and Shield Platform Encryption was rolled out over PHI fields.

Business change

Intake now runs on a single point of entry with the EHR and Salesforce in continuous sync, hardened across 16 iterative production deployments. We publish the architecture facts; no time-saved percentage was documented for this engagement.

Challenge

When sales marked an admission Closed Won, clinical staff still had to create the patient, open the episode of care, and file documents in the EHR by hand — and paper intake forms added another manual step before treatment could start.

System

GAT built a single Queueable that orchestrates 12 idempotent API callouts to the behavioral-health EHR on close — creating or updating the patient, opening the episode of care, and syncing documents — with token caching and environment switching held in custom metadata. Paper intake became a one-click PDF: a 48-field clinical screening rendered server-side and filed against the patient journey, plus inbound verification-of-benefits emails that attach themselves to the right Opportunity.

Business change

Sales sees revenue; clinical sees a ready patient record — the handoff between the two no longer depends on re-keying. We publish the pipeline's architecture; no cycle-time reduction figure was documented for this engagement.

Challenge

Earlier admissions, later discharges, exclusions, and inter-facility transfers each broke the census differently, and staff were reconciling the number by hand while also running an admissions funnel on paper-adjacent steps.

System

GAT decomposed the problem into a declarative recalculation engine — one main flow and six subflows covering earlier and later admissions, discharges, exclusions, and inter-facility transfers, roughly 1,630 lines of Flow logic — delivered with six architecture briefs so the client's team owns it. Around it, GAT built the admissions funnel itself: pre-assessment, verification of benefits, and ops handoff, with record-type-aware clinical PDFs generated server-side and the SMS console staff already used embedded directly on record pages.

Business change

The census now recalculates itself when dates move, and the admissions funnel runs in Salesforce end to end. The published figure measures delivery scale; no error-rate delta was documented for this engagement.

Challenge

Intake, consent, and clinical documents were being assembled from information already held in Salesforce, creating repetitive work and opportunities for inconsistent paperwork.

System

GAT mapped Salesforce record fields into more than 100 Docrio templates, organized the template set around the covered workflows, and made document generation part of the record process rather than a separate manual task.

Business change

Teams could generate the covered document set from Salesforce data instead of rebuilding each document by hand. We do not publish a time-saved or error-reduction percentage because one was not documented for this engagement.

Challenge

The org carried 216 flows, 45 custom objects, and years of undocumented decisions — with live revenue running through it and no sandbox to experiment in. Changing anything blind risked attribution, billing, and routing at once.

System

GAT retrieved roughly 5,200 metadata files and produced a version-controlled org map — attribution, billing, routing, and enrichment subsystems documented section by section, with AI-assisted analysis accelerating the archaeology. The forensic pass untangled two parallel attribution engines and two coexisting round-robin routing systems writing to the same records, caught live buttons wired to obsolete flows, and mapped the full HubSpot-to-Salesforce funnel.

Business change

The client got a de-risked org and a durable source of truth: every subsequent change is made against documentation instead of tribal memory. The published figures measure audit scope, not a revenue outcome.

Challenge

Loan disbursements had to leave Salesforce as correctly formatted, encrypted FedWire and NACHA bank files, reach a top-tier U.S. bank's gateway, and come back as acknowledgment files that close the loop on every transaction — all inside governor limits.

System

GAT built the payments infrastructure: batch jobs that assemble bank-format files, encrypt them with AES-256-CBC, and transmit them to the bank's gateway; inbound REST endpoints that ingest acknowledgment files and reconcile every transaction's status; and a GL-to-Oracle export pipeline for the finance team. Everything shipped with paired test classes, end-to-end documentation, and QA scripts through formal pull-request review.

Business change

The firm moves wires and ACH from inside its system of record with a reconciled status trail on every transaction. We publish the architecture and rigor; transaction volumes and counterparties remain confidential.

Challenge

The rental platform's API covers reservations, fleet, payments, and telematics — but handing an AI assistant raw write access to a live business is how demos become incidents. The client needed useful surfaces with hard guardrails.

System

GAT built a 114-tool MCP server over the rental platform's API, then carved two safe surfaces from it: a public ChatGPT app exposing exactly three read-only shopping tools behind strict output allowlists, and a private staff connector that annotates every tool, blocks writes by default, and gates payments, refunds, and deletes behind a second high-risk switch.

Business change

The system runs in daily operations today, with zero-diff schema parity verified after a full SDK rebuild. The published figure measures the tool surface delivered, not a revenue outcome.

Challenge

Salesforce teams ask the same expensive questions — what does this org do, what breaks if we change it — and the answers live in metadata no one reads. GAT built the product it wished every client engagement started with.

System

A Python backend orchestrates 12+ specialized agents over a pgvector RAG corpus of 92,000+ metadata chunks, and an MCP gateway lets Claude read, stage, and deploy against a live org with guardrails and per-org access control — with audit logging and PHI-redaction controls built in as the product works toward HIPAA compliance.

Business change

As of July 2026, ApexGenius has delivered 16,550 AI answers, connected 47 Salesforce orgs, and metered 2,645 MCP tool calls. These are product-telemetry figures from our own database, published as adoption measures — not client revenue claims.