One Unified Platform for a Regulated Trust Company: How Loop8 Built an Agentic AI Platform for Heritage Trust Company on Amazon Bedrock AgentCore
Heritage Trust Company partnered with Loop8 LLC to unify matter and client information fragmented across four SaaS systems and consolidate a previously separate sanction-screening application onto a single platform — turning hours of manual cross-system reconciliation into seconds of source-cited answers, with all data at rest in Canada.
Active matters unified
Estate and trust matters answerable from one plain-language interface
Logins consolidated
Four SaaS systems plus a standalone screener behind a single sign-on
To a working platform
Live production data with real credentials in Heritage Trust's own AWS account
Source-cited answers
Every model output verified against a real source record before display
Executive Summary
Heritage Trust Company, a trust and estate administration firm operating under strict regulatory oversight, had matter and client information fragmented across four SaaS systems that did not talk to each other, and a fifth application — the sanction-screening tool — running on separate infrastructure with its own login. Answering an ordinary question — “what is the status of this estate?” — required staff to sign into each system separately, reconcile the answers by hand, and then jump to another application for sanction screening. The reconciliation itself was the work.
Loop8 LLC, an AWS Consulting Partner, designed and delivered the Heritage Trust AI Integration Platform on Amazon Web Services in Canada. The four-week engagement produced four capabilities in one unified platform: an agentic Global Matter Search built on Amazon Bedrock AgentCore, a scheduled quarterly report draft generator, a monthly AML client-name export that never touches the LLM, and — as a follow-on to the original scope — the migration of Heritage Trust's sanction-screening application from a separate third-party hosting environment onto the same platform, giving staff a single sign-on, single interface for every AI-assisted workflow they touch.
About the Customer
Heritage Trust Company is a trust and estate administration firm that administers estates, trusts, and client matters in a highly regulated industry. Its practice touches four categories of information every day — matter records and ledgers, correspondence and documents, accounting, and estate- administration workflows — each living in a separate SaaS system. Sanction screening — a required compliance workflow in the trust and estate industry — ran on its own application entirely. Heritage Trust's obligation to clients and regulators is to answer questions across every one of those systems accurately, quickly, and with the source of every fact traceable back to the original record.
Background
Before engaging Loop8, Heritage Trust's day-to-day operations depended on staff moving between four SaaS tenants and reconciling their outputs by hand, with a separate sanction-screening application running independently on top of that. Answering a matter question meant signing into each source system, gathering related records, building a picture by hand, and then switching context to the screener for the compliance side of the workflow.
The picture was often already out of date by the time it was written down, and reconstructing which record each fact came from — for a beneficiary, counsel, or a regulator — was itself another manual exercise.
The Challenge
Manual cross-system reconciliation cost Heritage Trust in three places at once. First, it cost time: every matter question required the same repeated assembly across four systems, and every compliance action required a context switch to yet another. Second, it cost traceability: the source of each fact lived in the staff member's memory of where they had just looked, rather than in a citation attached to the answer. Third, it cost defensibility: in a highly regulated industry with a chain-of-custody obligation on both records and operational actions, reconstructing that chain from four fragmented audit logs plus a separate screener was neither fast nor auditable.
Two structured workflows compounded the problem. Each month, staff assembled a deduplicated active-client list from Clio and Estateably for the sanction screener. Each quarter, staff drafted status and financial reports per matter from scratch. Both workflows were mechanically tractable but manually onerous — and both were candidates for automation the moment the four systems could be read from one place.
There was also a compliance constraint that shaped everything. Client data had to stay resident in Canada, and any use of a large language model in the workflow had to be documented explicitly enough to satisfy the firm's internal governance and its regulatory environment.
Why Loop8
Heritage Trust selected Loop8 for the combination of SMB focus, full-stack delivery, and regulated-industry experience without regulated-industry overhead. Loop8 delivered the complete solution — infrastructure-as-code, the agent stack, the SharePoint knowledge base, the scheduled quarterly and AML workflows, the React front end, SSO federation, the compliance audit trail, and the technical review package prepared for the firm's internal governance review — as a single integrated engagement rather than requiring multiple vendors.
Loop8 designed the platform for a highly regulated industry from the first line of Terraform: read-only across every source system, PII masked before entering the prompt path, citation verification on every model output, and immutable audit stores.
“An AI partner that takes compliance seriously rather than treating it as friction.”— Nicole Garton, President & General Counsel, Heritage Trust Company
Why AWS
Amazon Bedrock AgentCore provides the managed control plane for the specific shape of this workload — long fan-outs across multiple SaaS APIs, per-session isolation, credentials that rotate cleanly under concurrent access, and conversational context that survives across turns. Building the equivalent on hand-managed infrastructure would have required assembling those primitives independently, each a new operational risk.
Model Context Protocol (MCP) is an open standard, which means the tool surface exposing Heritage Trust's four source systems is not coupled to a specific model, framework, or vendor. Amazon Bedrock (Claude) provides managed access to the foundation model for synthesis, governed by the AWS Data Processing Addendum with zero data retention — Heritage Trust did not need to fine-tune a model, manage GPU capacity, or host an LLM in a regulated environment. Amazon Bedrock Knowledge Bases with Amazon S3 Vectors as the vector store made it possible to run RAG over the SharePoint document corpus entirely in Canada.
Data residency in Canada is a first-class property of the platform: every component runs in Canada. And AWS gave Heritage Trust the primitives to build a chain of custody an examiner can rely on — an append-only activity log, control-plane audit through AWS CloudTrail, and immutable storage of every generated document in Amazon S3 with Object Lock.
The Solution
Loop8 designed and delivered the Heritage Trust AI Integration Platform in a four-week engagement, followed by a workstream that migrated the standalone sanction-screening application onto the same platform. The result is one place for staff to sign in, one interface for every AI-assisted workflow, and one audit trail behind them all.
Global Matter Search Agent
The flagship capability is a plain-language matter search agent. A staff member asks a question about any matter; the agent reads across Clio, Microsoft 365 / SharePoint, QuickBooks Online, and Estateably; and returns a source-cited status summary — current status, outstanding items, correspondence, tax and KYC status, upcoming deadlines, risks, and a suggested next action — with every assertion linked back to the specific record it came from. The agent is built on Amazon Bedrock AgentCore, uses Model Context Protocol (MCP) for interoperability, and calls Amazon Bedrock (Claude) for the final synthesis step. Document retrieval is served by Amazon Bedrock Knowledge Bases with Amazon S3 Vectors.
Scheduled Compliance Workflows
A quarterly report draft generator produces per-matter status and financial report drafts into a SharePoint review folder for a person to refine and send. A monthly AML client-name export compiles the deduplicated active-client list from Clio and Estateably into a formatted workbook for use with the sanction screener. Both run on an Amazon EventBridge schedule using AWS Lambda. The AML export performs structured extraction only — no client PII is ever sent to the LLM.
“By the final run of the AML export, every figure in the workbook reconciled line by line. That matters when the output supports a sanctions screening obligation.”— Nicole Garton, President & General Counsel, Heritage Trust Company
Sanction Screener Migration
The sanction-screening application, previously running on separate third-party hosting infrastructure with its own login, was migrated onto the same AWS platform Loop8 built for Heritage Trust. Staff now sign into a single interface for matter search, quarterly reports, AML export, and sanction screening — one login, one interface, one audit trail, one data-residency posture.
Web Application, SSO, and Data Residency
The staff-facing web app is a React single-page app on Amazon S3 and Amazon CloudFront. Staff sign in with their existing Microsoft 365 credentials via Amazon Cognito federated to Microsoft Entra ID with MFA enforced. Every component of the platform runs in Canada, so all data at rest stays there; the single cross-border flow — the Claude synthesis call under the AWS Data Processing Addendum with zero data retention — was documented explicitly for the firm's internal governance review.
Responsible AI and Immutable Audit
Responsible AI is designed in from day one: the platform is read-only across every source system, PII is masked before content ever reaches the model, every citation the model emits is verified against a real source record before it is displayed, and every external output remains human-in-the-loop. The compliance chain of custody lives in an immutable audit trail — an append-only activity log in Amazon DynamoDB, control-plane coverage in AWS CloudTrail, and immutable storage of every generated document in Amazon S3 with Object Lock.
Infrastructure as Code
The entire infrastructure is defined in Terraform and deployed by GitHub Actions authenticating to AWS via OpenID Connect federation — no long-lived AWS credentials exist in the pipeline. The platform runs in Heritage Trust's own AWS account, with all source code, infrastructure-as-code, and operational documentation transferred at production cutover.
Results
Heritage Trust administers more than 200 active estate and trust matters, and each of them used to require the same repeated cross-system assembly. Where a matter question used to demand signing into four separate systems and reconciling their outputs by hand, the same question now returns a source-cited status file in seconds. Every assertion carries a citation that resolves to a specific record in a specific system, so beneficiaries, counsel, and regulators can be shown exactly which record every fact came from.
The compliance posture changed at the same time. Data residency is no longer fragmented across four SaaS tenants — every piece of client data at rest resides in Canada, and the one cross-border flow is documented explicitly rather than assumed. The audit trail is no longer four separate SaaS logs — it is a unified, append-only activity log, control-plane coverage in CloudTrail, and immutable storage of every generated document. And by migrating the sanction-screening application onto the same platform, Heritage Trust consolidated a fifth compliance-sensitive workflow behind the same audit trail, the same data-residency posture, and the same single sign-on.
The scheduled workflows now run themselves. The Monthly AML Client-Name Export compiles and deposits the deduplicated active-client list on the 1st of each month without staff involvement, with every figure in the workbook reconciling line by line against the source systems. The Quarterly Report Drafts arrive in the SharePoint review folder ready for a person to refine and send — instead of a blank page to write from scratch. Heritage Trust and Loop8 are already in discussions on ongoing support and a next phase.
Operational Outcomes
- Matters unified200+
- Logins consolidated4 → 1
- Answer time per matterSeconds
- Source citations100%
Platform Foundation
- Agent runtimeBedrock AgentCore
- InteroperabilityMCP
- Data residencyCanada
- Pricing modelPay-per-use
— Nicole Garton, President & General Counsel, Heritage Trust Company“Heritage Trust administers more than 200 active estate and trust matters, and our information lived in four systems that did not talk to each other. Loop8 built us a platform on AWS that unifies them, with read-only connectors to Clio, Microsoft 365, Estateably and QuickBooks, single sign-on through Entra ID, a plain-language matter search agent, automated quarterly report drafts and a monthly AML name export.
What made the engagement work was not just the build. It was how the team handled testing. We tested exhaustively and sent back detailed, often inconvenient lists of problems. Loop8 treated every item as worth fixing rather than worth arguing about, and came back with fixes deployed and a written memo mapping each one to what had changed. By the final run of the AML export, every figure in the workbook reconciled line by line. That matters when the output supports a sanctions screening obligation.
We would recommend Loop8 without hesitation, particularly to firms in regulated sectors who need an AI partner that takes compliance seriously rather than treating it as friction. We are already discussing ongoing support and a next phase.”
AWS Services Used
About the Partner
Loop8 LLC is an AWS Consulting Partner that delivers production-grade cloud and AI solutions to small and medium businesses. Loop8 holds the AWS SMB Competency and specializes in AI and cloud business and technology transformation — building enterprise-grade AI agentic tools and workflows on AWS, delivered at small-business speed and price. Loop8 helps customers move from idea to working platform in weeks, not quarters — bringing the engineering depth of a larger consultancy to customers who do not have the time or budget for a multi-vendor, multi-quarter program.
Bring Agentic AI to Your Regulated Workflow
See how Loop8 can consolidate fragmented systems into a single AI-powered platform — with compliance treated as a design constraint, not an afterthought.