BYRON 3.0Byron moved the margin. Byron 3.0 scales the engine.
Byron carries the founder's name.
Named for founder Byron Laflin Smith. Now firmwide across sales, analytics, programming, and fraud prevention — not a pilot.
of the 2025 expense base, saved by AI.
NT Byron's productivity savings exceeded 4% of the 2025 expense base. Every employee now has access.
margin target — up from 29.2% in 2025.
The AI productivity engine is NT's largest lever to close the ~400bps gap. Byron 3.0 keeps it compounding.
in assets under custody & administration.
NT's stated AI strategy names infinite scalability as its third outcome. Byron 3.0 is the platform layer that governs and scales agents to that mandate.
Retrieval
SharePoint-connected agents. Users had to know which agent to ask.
Orchestration
A parent agent routes intent to child agents. Built in Copilot Studio.
Enterprise intelligence
One governed layer across every framework, vendor, and business unit.
Byron 2.0 proved the model at Northern Trust's scale. Byron 3.0 makes it structural — the enterprise intelligence layer beneath every agent, in every framework, across the firm.
Byron's success created three new realities.
Demand is expanding.
Business units increasingly want their own agents. Intake runs at 1–2 new requests per week, projected to exceed 50 business-unit agents by Q4 2026.
Consequence: the per-agent delivery model breaks.
Development is diversifying.
Agents are no longer Copilot Studio alone. Other pro-code frameworks (LangChain, LangGraph), digital coworkers, and third-party vendor AI now coexist.
Consequence: governance fragments across frameworks.
Governance stays uncompromising.
As adoption accelerates, enterprise expectations don't soften. Model risk, security, privacy, and compliance review must hold at custodian scale.
Consequence: without a platform, innovation slows.
Northern Trust's challenge is no longer building intelligent agents. It is institutionalizing the operating model that lets hundreds of agents be delivered with the same governance, trust, and efficiency as the first.
Every agent, built from scratch.
Every agent inherits the platform.
Grow AI adoption across business units without multiplying operational complexity.
Coordinate agents, systems, and knowledge across a heterogeneous AI ecosystem.
Deliver secure, explainable, observable AI that users and control functions can rely on.
Policy, compliance, and lifecycle as platform properties — routed through NT's own review bodies.
Four capabilities, implemented as one layer.
↑ inherited (Byron 2.0)
Metadata standards · content hygiene · orphaned-library archiving · RAG on NT content.
Available & upcoming agents by function. Discover, use, express interest.
Prebuilt workbenches per business function + custom workflows.
★ primary
OPTIMUSOn-prem agent appliance
with open-source models
Northern Trust has identified low maturity across metadata, content lifecycle, and semantic modeling — Byron 3.0 is the firm's first structured enterprise knowledge layer. This is where the platform earns retrieval trust as content and agents multiply.
Fail at the evaluation gate — automatic rollback, environment cleaned up, nothing reaches production.
Byron 2.0's Copilot Studio orchestration keeps running. Pro-code frameworks keep building. Lyzr adds the enterprise layer both need to move from scattered agents to a governed estate — without rebuilding what already works.
Copilot Studio, LangChain / LangGraph — build agents natively
Registry every agent onboards into — one identity model, one governance gate, regardless of who built it or where
Byron 2.0 parent agent routes across Copilot Studio children
Cross-framework routing — intent routing and handoffs across Copilot Studio, other pro-code frameworks, and third-party AI
SharePoint, ServiceNow, Power BI connectors exist as native integrations
Reusable connector patterns & MCP — one integration pattern that scales across business units instead of bespoke per agent
Framework-level guardrails on individual agents
Platform evaluation gate — accuracy, relevance, and policy checks before any agent reaches production traffic
Logging and access controls within each framework's own environment
One control plane — one entitlement model, one audit trail across every framework, routed through NT's review bodies
Byron 2.0's agents aren't thrown away — they become the first tenants of the layer. Every agent after them inherits governance, identity, evaluation, and connectors already built. The estate Northern Trust already runs becomes its foundation.
One platform. Three enablers on top.
Lyzr Studio
The platform that powers all three enablers — build, run, govern, and improve agents in one place, inside Northern Trust's own cloud boundary.
Consulting
Use-case identification and prioritization, journey planning, value realization, and MBR / QBR frameworks — the top-down view of where Byron 3.0 creates the most value.
Forward-Deployed Engineers
Applied engineers embedded with Northern Trust's team. They accelerate the first builds, then transfer capability so NT engineers own the platform.
Training + Knowledge Repository
Enablement for NT independence. Reusable agent patterns, reference architectures, and deployment playbooks — so the second, tenth, and hundredth agent onboard faster than the first.
Governance built in
Registry, identity, evaluation, audit trails, and human-in-the-loop are platform capabilities from day one — not 5–6 months of engineering before the first workflow ships.
Orchestrate, don't migrate
Existing Copilot Studio and pro-code agents connect as-is. This matches NT's own stated near-term strategy and removes the single largest cost driver in most platform programs.
Capability transfer, not lock-in
Code-level IP transfer of accelerators, connectors, and governance workflows into NT's own repository. The platform becomes Northern Trust's operating capability, with Lyzr in an advisory role.
Proven where trust must be earned.
Global consulting firms. Fortune 500 enterprises. Regulated financial institutions. Lyzr has already been chosen where the stakes are highest and the scrutiny is deepest.
Trusted by the environments.
Strategic investors betting on Lyzr's agentic future.
Awards and analyst recognition.
By Mosaic Score. Top 6% of the cohort.
Named vendor in the emerging agentic AI category.
Agentic Technology category.
Recognized in the leading independent evaluation.
The credentials that matter.
Northern Trust brings 136 years of institutional trust and the Byron platform it has already proven. Lyzr brings the enterprise layer to scale it. Together — a governed intelligence platform that compounds with every agent, every cycle.
Four phases. Lyzr's role shrinks at each one.
Define
Byron 3.0 architecture, orchestration, identity and security model, platform decisioning.
Design
Agent interaction patterns, routing, handoff protocols, governance and intake model.
Build
Intelligence-layer MVP, first agent cohort registered and evaluated, security validated. NT engineers shadow, then co-own.
Scale
Operating model, registry scale-up, hypercare — then capability handed fully to NT teams. Lyzr shifts to advisory and support.
Enterprise Service Levels & Resilience
- Platform availability aligned to enterprise usage
- Incident classification & time-bound response by impact tier
- Orchestration latency and throttling to protect core use cases during peak
- Regular reporting on platform health, routing outcomes, and policy events
- High availability and disaster-recovery alignment
- Graceful degradation — individual agent or integration failures don't take down the platform
- Predictable behavior when context, permissions, or downstream services are unavailable
- Artifacts provided for audit, risk, and compliance reviews
Byron 3.0 is the foundation, not the destination. The same governed layer extends to an enterprise agent marketplace, cross-business intelligence, and — on Northern Trust's own roadmap — client-facing AI services.
One number. Shared risk.
- Full Lyzr platform — governance, orchestration, lifecycle, evaluation, environments
- Delivery team funded by Lyzr — 7 people, no charge
- On-site FDE consultant, 6 months
- Bangalore agentic pod — 6 resources, 12 months
- No separate charge for implementation, integration, or usage
- 200 billion TPL included in base
- $250,000 per additional 50 billion TPL — flat, no step-ups
- 250B → $1.25M · 300B → $1.5M · 400B → $2.0M
- Delivery support optional — NT teams, Lyzr pod, or hybrid
| Year One | Year Two | |
|---|---|---|
| Platform license | $1,000,000 | $1,000,000 base |
| Usage | Unlimited | 200B TPL included |
| Additional consumption | None | $250K per 50B TPL |
| Delivery & engineering | Included · Lyzr-funded | Optional |
| LLM inference & compute | NT direct | NT direct |
| Agentic IP ownership | Northern Trust | Northern Trust |
- LLM inference — NT's own model providers, no Lyzr markup
- Cloud & hosting compute — NT's own Azure
- Model- and cloud-agnostic, change anytime
- NT owns everything built — agents, prompts, knowledge, integrations, Byron 3.0 in full
- Open-source core, no proprietary runtime dependency
- Exit assistance & open-format export committed at signature
The RFP invites both fixed-bid and T&M. Lyzr submits a fixed bid only — T&M pays a vendor more when delivery takes longer, pointing the incentive the wrong way on an intelligence layer. Lyzr carries the schedule risk instead. A T&M schedule is available on request.
Byron 3.0 creates value when it reaches scaled, governed production — not when a timesheet is filed. Lyzr funds the path there and prices the platform as one fixed, defensible number.