CBOS

If your systems are making work harder than it needs to be, get in touch with CBOS.

𝐓𝐡𝐞 𝐅𝐮𝐭𝐮𝐫𝐞 𝐨𝐟 𝐓𝐫𝐚𝐢𝐧𝐢𝐧𝐠.
Training has always existed to grow a company’s capabilities and worth. As of late, though, training has gained the tag of a grudge purchase, or it is ignored completely. Most companies, however, don’t have a training problem; they have a training production problem.

The knowledge already exists somewhere, whether in policies, SOPs, product documents, safety requirements, sales material, onboarding packs, and the heads of experienced staff.

The problem is turning that knowledge into usable training quickly enough.

Traditional training systems still assume that course creation is slow, expensive, and separate from the business. You write the content, design the course, upload the material, assign it manually, chase completions, and then try to work out whether anything actually changed.

Y was built to change that.

Y uses AI to help companies generate training courses faster, update them as the business changes, and connect training back into the systems that already run within the company.

That means you can create your courses faster, for a lower development cost, while having complete visibility over your training and compliance. There’s less admin, and training can actually keep up with products, processes, and people instead of being months behind.

But the real value is not just that Y creates courses.

The value is that Y turns training into an operating system.

A way to ask:
Who has been trained?
Where are the gaps?
Which branch is least compliant?
What needs to be updated?
What is falling behind?

Training should not be a once-off event that disappears into an LMS.

It should be live, connected, and visible.

That is what Y is built for.

Join our pre-release waitlist for 10% off your first month: y-institute.com

The reality is that if you are utilising multiple different applications in separate departments, we can guarantee you are losing money and wasting resources. Unfortunately, this compounds for each instance where a human is doing the work the system should have been designed to fulfil.

You shouldn't have to wait for someone to send you data; you should have immediate access to it at all times. You are losing out to your competitors if you haven't streamlined your systems, because you're wasting the time of your best employees by asking them to do menial tasks that don't contribute to your company's success.

If you're interested in understanding a little more of what we've noticed around this over the years, read our latest article.

Connect your systems before it’s too late. Your biggest loss in business is seldom shown on your invoice.

𝗦𝗼 𝗠𝗮𝗻𝘆 𝗙𝗿𝗮𝗺𝗲𝘄𝗼𝗿𝗸𝘀... 𝗪𝗵𝘆 𝗖𝗵𝗼𝗼𝘀𝗲 .𝗡𝗘𝗧?

Choosing the right framework is never simply about performance benchmarks or popularity charts; it is about stability, maintainability, ecosystem maturity, and long-term support. Python frameworks such as Django, Flask, and FastAPI each offer compelling strengths, but they come with trade-offs in asynchronous consistency, structural enforcement, or ecosystem maturity. Node.js provides enormous flexibility and an unmatched package ecosystem, yet places structural responsibility squarely on the developer. Go offers performance and simplicity through compilation and concurrency control, though certain language features and patterns are still evolving. Java with Spring delivers enterprise-grade robustness and structure, but requires careful management of the JVM ecosystem and configuration complexity.

This is where C # and .NET distinguish themselves. Unlike many language–framework combinations where versioning, governance, and ecosystem direction are fragmented, .NET operates as a unified platform under a clear and consistent release strategy. The alignment between the C # language and the .NET runtime reduces the uncertainty around compatibility. The predictable cadence of short-term and long-term support releases provides both innovation and stability.

Beyond governance considerations, .NET offers a well-rounded architecture: near-native performance, mature asynchronous capabilities, clear and enforced project structure, built-in security, and strong first-party integrations that minimise reliance on third-party dependencies. Developers can choose minimal APIs for lightweight services or structured controllers for enterprise applications, without sacrificing cohesion across the platform.

While no framework is without drawbacks, .NET’s blend of performance, structure, tooling, and long-term stewardship makes it a compelling choice for modern API development. It offers not just a way to build applications, but a stable and scalable foundation upon which those applications can evolve confidently over time.

Damian Matthews, a senior developer at CBOS, pulls together all these points in his full article:

So many frameworks… Why you should choose .NET?

The concept of "context" in AI is often reduced to "token window size," but its true importance is philosophical and structural. In the world of Large Language Models (LLMs), data provides the bricks, but context provides the architecture. Without it, AI is just a highly sophisticated autocomplete.

𝟭. 𝗧𝗵𝗲 𝗗𝗲𝗮𝘁𝗵 𝗼𝗳 𝘁𝗵𝗲 "𝗔𝘃𝗲𝗿𝗮𝗴𝗲" 𝗥𝗲𝘀𝗽𝗼𝗻𝘀𝗲
Without context, an AI defaults to the statistical mean of its training data. If you ask for a "good workout plan" without providing context, the AI gives you a generic 3-day split because that is the most common answer in its database.

Context allows the AI to collapse the wave function of infinite possibilities into a single, high-fidelity reality. It transforms the model from a generalist librarian into a specialized consultant.

𝟮. 𝗦𝗼𝗹𝘃𝗶𝗻𝗴 𝘁𝗵𝗲 "𝗜𝗻𝘁𝗲𝗻𝘁 𝗚𝗮𝗽"
Human communication is notoriously ambiguous. We rarely say exactly what we mean because we rely on shared context (history, tone, environment) to fill in the gaps.

· No Context: "Add some AI features to our platform."

· With Context: "Build an AI workflow that automates our specific support escalation process, using our unique customer history and internal resolution playbook."

Context is the bridge that spans the gap between what a user types and what they actually need.

𝟯. 𝗗𝘆𝗻𝗮𝗺𝗶𝗰 𝗥𝗲𝗮𝘀𝗼𝗻𝗶𝗻𝗴
We are moving away from the era of "Knowledge Retrieval" (where AI acts like Google) and into the era of "In-Context Learning." Modern AI doesn't necessarily need to be "retrained" on new data to understand it. If you feed a model 50 pages of your specific project notes, it "learns" those notes within the session. This makes context the operating system of the interaction. The model provides the processing power, but your context provides the software it’s currently running.

𝟰. 𝗧𝗵𝗲 "𝗔𝗻𝗰𝗵𝗼𝗿" 𝗔𝗴𝗮𝗶𝗻𝘀𝘁 𝗛𝗮𝗹𝗹𝘂𝗰𝗶𝗻𝗮𝘁𝗶𝗼𝗻
Hallucinations often happen because the AI loses the thread of the "truth" and starts following the strongest statistical path of words.

Context acts as a grounding wire. By providing a "Source of Truth" (like a document or a specific set of rules), you constrain the AI's creativity within the bounds of reality.

𝗧𝗵𝗲 "𝗦𝘁𝗮𝗸𝗲𝘀" 𝗼𝗳 𝗖𝗼𝗻𝘁𝗲𝘅𝘁
As we move toward Agentic AI, context becomes a matter of safety and utility. An AI agent tasked with "managing a company system" is useless, and potentially dangerous, if it doesn't have the context of your business preferences, your clients, your employees, and various other factors.

Context is the architectural constraint that transforms a probabilistic engine into a functional tool.

𝟭𝟬 𝗧𝗶𝗽𝘀 𝘁𝗼 𝗞𝗲𝗲𝗽 𝗮 𝗖𝗹𝗶𝗲𝗻𝘁: 𝗙𝗿𝗼𝗺 𝗮 𝗖𝗘𝗢

Winning a project, whether the easy or hard part, will never compare to the work required to maintain the relationship with a client.

In technology, relationships often start with the flashy features and proposals, but long-term partnerships are built on something far simpler: understanding the client’s business deeply enough to care about the outcome, not just the implementation.

Listening before selling.
Setting realistic expectations.
Staying engaged long after go-live.

Basically, not overpromising and then underdelivering, as a way of landing the deal in the first place.

These are small behaviours that compound into trust.

Over time, the role of a technology partner changes. You are no longer just a provider, but also the insight into the decisions around their business (efficiency, risk, growth, and strategy). That’s when a supplier becomes part of the conversation long before the next project is defined.

Here are 10 practical principles for building long-term client relationships from a CEO’s perspective.

In more than a decade of operation, we have come across multiple industries. Here are five of the industries we have made a big impact in.

𝐅𝐮𝐧𝐜𝐭𝐢𝐨𝐧𝐚𝐥 𝐀𝐠𝐞𝐧𝐭𝐢𝐜 𝐀𝐈
For decades, we’ve operated on the bedrock of determinism: If you give the machine the same input, you get the same output. Our entire SDLC, from unit tests to sprint velocities, was built to enforce this mechanical certainty.

But with Agentic AI, that concept is being torn up.

We are moving away from systems that simply execute logic and toward systems that pursue goals. Agentic AI is not just a "smarter" layer of code. It represents a fundamental shift in how software behaves.

When software becomes probabilistic, our mandate shifts from managing the logic to governing conduct. Success is no longer measured by whether the code ran, but by whether the system’s behaviour remained aligned with our intent.

How we are evolving the SDLC for an Agentic world:
• From Features to Intent: Planning is now about constraint engineering. We must be as precise about the guardrails of what an agent cannot do as we are about the objectives it should achieve.

• Scaffolding: We aren't building architecture in the same way anymore. It isn’t a definitive guide but rather a containment field for agentic AI to function within. This requires a new layer of validator agents and memory structures to ensure autonomous actions stay safe and observable.

• The End of Binary Testing: The "Pass/Fail" unit test is insufficient for an agent. We are shifting toward behavioural evaluation, measuring alignment, risk, and drift rather than just checking for crashes.

The transition from the mechanical to the behavioural is the next great engineering challenge. It requires us to stop trying to control every line of code and start learning how to govern the accountable autonomy of our systems.

Read our full analysis on the "Post-Deterministic SDLC" here:

Functioning Agentic AI Engineering Intent in the Age of Agentic AI

𝐓𝐡𝐞 𝐃𝐞𝐚𝐭𝐡 𝐨𝐟 𝐭𝐡𝐞 𝐓𝐢𝐦𝐞𝐬𝐡𝐞𝐞𝐭

As software absorbs more of the mechanical aspects of work, the value shifts upstream. What matters most is no longer how long something takes, but whether it positively affects uncertainty, risk, and behaviour within a company.

Time-based thinking made sense when work correlated well with each hour.
It makes far less sense in modern knowledge work.

This article explores why hours, utilisation, and effort are increasingly poor proxies for value and why outcomes are becoming the only metric that really holds.

Read it here:

The Death of the Timesheet For much of modern economic history, work has been priced according to time. Hours, rates, utilisation, and capacity have served as the…