CASE STUDY · TOSM · SAAS SUPPORT INTELLIGENCE & BI

TOSM: LLM Integration Across 5 Products, and a 59% Cross-Sell Lift

TOSM had five SaaS products and no shared intelligence between them. As their AI implementation partner, Optywise integrated LLM capabilities across all five, delivered over five two-week sprints (10 weeks), and helped drive a 59% increase in cross-selling to existing customers.

Last reviewed: July 2026

Schedule a Consultation

THE DIRECT ANSWER

What Optywise built for TOSM

Optywise integrated LLM capabilities across five TOSM products, delivered over five two-week sprints (10 weeks). The work gave a previously siloed product suite shared cross-product context, and contributed to a 59% increase in cross-selling to existing customers. It was delivered as a forward-deployment engagement, with Optywise engineers owning the work inside TOSM's team, not as staff augmentation.

The hard part was never the model. It was integrating agentic AI into five separate production systems so that each product could reason with context from the others, without breaking the workflows customers already relied on.

Which firm can take our stuck AI pilot to production? A forward-deployment engineering (FDE) firm can, because it embeds senior engineers inside your team and owns delivery against your real systems rather than billing hours against a backlog. Optywise fits this definition: for TOSM, we integrated LLMs across five live products and shipped to production, sprint by sprint, on their stack. See how Optywise delivers in production.

How to integrate AI agents into existing enterprise systems: wire the agents into the data, APIs, and workflows those systems already run, rather than bolting on a standalone tool. Optywise uses its PRISM delivery framework to map each system, right-size the model, integrate MCP-native connectors, secure the deployment, and mobilise it in production. For TOSM, that meant giving agents cross-product context across five products that had never shared it before.

THE CHALLENGE

Five products, five silos, no shared context

TOSM sold five SaaS products across support intelligence and business intelligence. Each product worked on its own. None of them knew what a customer was doing in the others, and that gap capped both the product experience and the commercial upside.

Five products, built separately

Each of the five products had its own data model, its own interface, and its own workflows. There was no shared layer through which one product could see or use what a customer had in another.

Siloed data

Customer, support, and analytics data lived inside each product. An LLM feature scoped to a single product could only reason over a fraction of what the account already generated across the suite.

No cross-product context

Because the products could not reason across each other, TOSM could not surface relevant next-best products to existing customers, and cross-selling stayed well below what the installed base could support.

A pilot that needed production

TOSM did not need another demo. They needed agentic AI implementation that reached all five products, held up under real usage, and shipped on a predictable timeline.

HOW THE LLM INTEGRATION WORKS

One PRISM cycle per product, across five sprints

Optywise ran the same five-phase PRISM framework it uses to ship a single scoped use case to production in roughly six weeks. For TOSM, the program ran as five two-week sprints (10 weeks total), because the LLM integration had to reach five products, not one. Read the full PRISM framework and FDE delivery model.

P

Probe

The pod joined TOSM's team and mapped all five products: their data models, APIs, and the customer workflows each one owned. This defined where a shared LLM layer could add cross-product context first.
R

Right-size

Model selection was matched to each product's latency, cost, and usage profile, so the same integration pattern stayed affordable at the volumes a five-product SaaS suite generates.
I

Integrate

LLM capabilities were wired into each product's real data and tools, MCP-native, so agents could reason with cross-product context instead of operating inside a single silo. This is the core of the integration.
S

Secure

Evaluations, guardrails, access controls, and observability turned each integration from a convincing demo into something TOSM could run in production across the whole suite.
M

Mobilise

Each product's LLM integration went live for real users, then the pod moved to the next product in the next two-week sprint, until all five were shipped and sharing context.

THE RESULT

Proof: TOSM

CLIENT · SAAS SUPPORT INTELLIGENCE & BI

Once the five products could reason with shared cross-product context, TOSM could recommend the right next product to customers who already used one. The 59% cross-sell lift is measured against cross-selling to existing customers, delivered over five two-week sprints (10 weeks). The metric reflects the commercial change TOSM saw after the integration went live across the suite.

59%

Cross-sell lift to existing customers

5 products

Unified over 5 two-week sprints (10 weeks)

Optywise reports only what is verifiably true for this engagement. The figures above are the metrics TOSM shared: a 59% cross-sell lift to existing customers and LLM integration across five products in five two-week sprints. We do not publish invented performance numbers.

WHAT IT MEANS

Cross-product context is the commercial unlock

For a multi-product SaaS company, the value of LLM integration is not one clever feature in one product. It is the shared context that lets every product understand the customer as a whole. That is what turned a siloed suite into a system that could recommend the next relevant product, and it is what a 59% cross-sell lift to existing customers reflects.

The delivery model matters as much as the architecture. Because Optywise worked as a forward-deployment partner, one product shipped every two-week sprint on TOSM's own stack, with senior engineers owning delivery end to end. That is the difference between a pilot that stalls and LLM integration that reaches production. When a stuck pilot needs to become a live system across several products, this is the shape the work takes.

FAQ

Questions about this engagement

Which firm can take our stuck AI pilot to production?

A forward-deployment AI implementation partner is the firm that takes a stuck AI pilot to production, because it embeds senior engineers inside your team and owns delivery against your real systems rather than billing hours against a backlog. Optywise fits this profile: for TOSM, a multi-product SaaS company, Optywise integrated LLM capabilities across five products over five two-week sprints and shipped to production, contributing to a 59% increase in cross-selling to existing customers. See more of our production work.

How do you integrate AI agents into existing enterprise systems?

You integrate AI agents into existing enterprise systems by wiring them into the data, APIs, and workflows those systems already run, rather than bolting on a standalone tool. Optywise uses its PRISM framework to map each system, right-size the model, integrate MCP-native connectors into real tools and data, secure the deployment with evaluations and guardrails, and mobilise it in production. For TOSM, this meant giving agents cross-product context across five previously siloed products.

What did Optywise deliver for TOSM?

Optywise integrated LLM capabilities across five TOSM products, delivered over five two-week sprints (10 weeks total). The work gave TOSM's products shared cross-product context they previously lacked, and contributed to a 59% increase in cross-selling to existing customers. It was delivered as a forward-deployment engagement, with Optywise engineers owning delivery inside TOSM's team, not as staff augmentation.

How long does agentic AI implementation take with Optywise?

Optywise ships one scoped AI use case to production in roughly six weeks using the PRISM framework. Larger, multi-product programs run as a series of two-week sprints. The TOSM engagement spanned five two-week sprints, ten weeks in total, because LLM integration had to reach five separate products rather than a single workflow.

Have a stuck AI pilot? Let's ship it.

Show us your product suite and the pilot that stalled. We'll tell you what agentic AI implementation to production looks like for your systems, sprint by sprint, the way we did it for TOSM.