Michael Jarry Aug 04, 2025

DevOps: The Structure that makes Innovation Possible

Listen to DevOps: The Structure that Makes Innovation Possible
21:31

Why DevOps Is the Backbone of Scalable Innovation

In the race to adopt transformative technologies like AI, cloud-native platforms and microservices, many organisations overlook the one thing that makes innovation sustainable: a strong operational foundation. 

DevOps has evolved from a rebellious mindset shift in how we approach software delivery into one of the de-facto IT management approaches that encompasses a set of practices to enable faster, safer and more scalable tech adoption. It’s not just about practices; it’s about the principles that drive them that impact “what we do” as much as “how we do them”.  

DevOps As A Strategic Enabler 

DevOps has been around for a while, but it is constantly evolving. It emerged from the recognition that operational silos between technical teams were fundamentally destructive and distracted from what really mattered: the outcome. At its core, DevOps is about creating shared ownership of the end-to-end delivery process, where everyone is responsible for delivering value to customers, instead of just being a set of checkpoints along the way.  

The real value of DevOps is in the shift in values. DevOps is about fostering collaboration, adaptability, accountability (without blame) and outcome-driven thinking. Whether you’re maintaining an existing product, building the latest AI startup, or somewhere in between, DevOps is toolset-agnostic and universally applicable. 

What a Strong Operational Foundation Looks Like 

A strong operational foundation enables innovation to scale safely and sustainably. In DevOps terms, this includes: 

  • Automation-first development and deployment: Reducing manual effort and human error by automating repetitive tasks such as builds, tests and deployments. 
  • Continuous integration and delivery (CI/CD): Ensuring that code changes are automatically tested and deployed, allowing for faster and safer releases. 
  • Observability across systems: Providing visibility into system behaviour through metrics, logs and traces, so teams can detect and resolve issues quickly. 
  • Governance that enables autonomy: Creating guardrails that allow teams to move fast without compromising compliance or security. 
  • Shared responsibility and collaboration: Most importantly, a shared responsibility for delivering business value across technical teams, driving collaboration, innovation and compromise.  

These foundational practices span multiple teams and are collaboratively implemented, with a focus on delivering what matters to the end customer. Without these elements, implemented in the right way, organisations often struggle to deliver innovation at pace or scale. 

DevOps Maturity Model 

Implementing a philosophy in concrete ways is difficult, because it will look slightly different to every organisation, but organisations can track their progress in achieving DevOps maturity by matching their approach to delivering change with one of these DevOps maturity stages: 

Ad Hoc  

Teams are reactive and putting out fires, with blame being handed out to those who drive change. Manual processes are used and teams operate to achieve their own metrics instead of delivering business value leads to stagnation.  

The marketing team urgently needs an update to the website for tomorrow’s product launch, which has been scheduled for months. The development team says they need two additional weeks because they need to get approval from security, coordinate with the operations team to organise the next deployment window, and have the testers sign off. When marketing escalates to management, they pressure the operations team to do an expedited change. The update introduces a bug that causes the website to intermittently crash, causing two hours of degraded service. In the post-incident meeting, the development team blames operations for the rushed deployment, operations blames the development team for not waiting until testing is signed off, and security points out that the change doesn’t align with process. Everyone agrees the change process is broken, but nobody takes ownership to fix it. 

Repeatable 

Basic processes are documented and repeatable. Version control and scripting are introduced, but automation is limited. Teams begin to reduce variability and improve consistency. 

The teams have met together, decided and documented their deployment process, and created a fortnightly change-advisory board. When the marketing team request another update, there’s now a clear process: submit a change request, get approval, and schedule it for the next deployment window in five days. The development team create a pull request, allowing the testing and security teams to review the code. Deployment goes smoothly because everyone knows their role, but the marketing team is frustrated by the lead time and the delay that it causes in rolling out customer-requested features. The developers know they could automate much of this process, but the business is cautious, focusing on consistency first. They’re now measuring deployment success rates and tracking performance impacts, even though deployment is slow. 

Automated 

Automation is applied to builds, tests, and deployments. CI/CD pipelines are established, enabling faster and safer releases. Teams start to measure performance and reliability. 

Marketing requests the update on Monday morning. The developers create a pull request for the new feature, which triggers automated unit and integration tests, and static and dynamic security scans. After code review, the change is automatically deployed to staging and a notification sent to the testing team with the deployment details. The testing team run an automated testing suite against the new change to validate functionality combined with some user acceptance testing. By Wednesday afternoon, marketing approves the change, and it’s automatically deployed to production during business hours with zero downtime using blue-green deployment and canary testing, progressively showing the new website to users and allowing for one-click rollback. The operations team monitor the deployment in real time through dashboards, prepared to roll back if any issue is detected. The whole process took two days with no manual intervention required, and all teams have total visibility of the change progress. 

Optimised 

Processes are continuously improved. Observability, governance, and feedback loops are embedded. Teams operate with autonomy, guided by metrics and aligned to business outcomes. 

The marketing team is currently testing campaigns and needs small, frequent updates to the website. Instead of requesting individual changes, they work with the platform team to build a self-service capability that lets them make direct updates to the website while staying within existing security and operational guardrails. When an update causes a slight increase in page-load times, the system automatically notifies the team, providing telemetry and summarisation of the issue and of the release that caused the increase. They quickly identify and fix the issue and add a new test to the automated testing pipeline to test for the issue in future. The platform team now focus on providing the business with self-service tools and implementing feedback to enhance existing toolsets and allow the business to innovate and experiment, adapting to market demands more safely and quickly. 

How to Assess Your Current State 

To determine where your organisation sits on the maturity curve, ask: 

Are our processes documented and repeatable? 
  • Are all teams aware of the process, and is it followed consistently?  
  • When something goes wrong, does blame fall to people, the process or to its rightful recipient, DNS?  
  • How do we handle exceptions when they arise?
  • Do the processes streamline the delivery of changes or hinder it?

How much of our delivery pipeline is automated? 

  • How much manual intervention does the change process require?  
  • What’s the business confidence about the impact of new changes?  
  • When there’s a failure, how long does it take to roll back?  
  • Can a minor code change be deployed safely and quickly?  

Do we have visibility into system performance, security, reliability and customer feedback? 
  • Do you find out about new problems through alerts or customer complaints?  
  • How easily can you trace a customer issue back to a specific code change?  
  • Can business stakeholders trust your system metrics enough to make decisions using them?  
  • What’s the length of time between an issue being introduced to the system and you finding out about it?  

Are teams empowered to make decisions and improve continuously? 

  • Can teams safely experiment with new ideas without facing scrutiny?  
  • When incidents happen, do you focus on the “who” or the “why”?  
  • Can teams access the data and tools they need to do their jobs effectively?  
  • Do teams regularly suggest and implement improvements to their workflows?  

Are we measuring meaningful outcomes? 

  • Is team performance measured by tickets closed or value delivered?  
  • Can stakeholders tell how long a feature will take to implement?  
  • Do technical teams (e.g. development and operations) share the same success metrics?  
  • When you optimise one metric, do others get worse?  

Does your organisation exhibit any of these red flag indicators? 
  • A “that’s not my job” mentality between teams 
  • Most changes are urgently pushed through or improperly planned 
  • A general fear of making changes 
  • Success depends on a few key special individuals 
  • Nobody can explain how changes connect to customer value 
  • Teams get blamed for failed changes instead of reflecting on the process 

Use this model to identify gaps and prioritise improvements that align with your strategic goals. 

Cultural Transformation 

Strong technological transformation fails without cultural change, even with the best tooling and standards.  Tooling should be used to achieve DevOps outcomes and standards used to define consistent processes, but without the freedom and trust to innovate, it’s impossible to use them to build anything meaningful. 

Building the right culture means:  

  • Shared ownership: Development teams, operations teams and business stakeholders work toward common goals, not individual department metrics. These goals need to come from the top, be measurable, and be clearly defined so that every team understands their contribution.  
  • Psychological safety: Growth, experimentation, and cross-team work is encouraged. When incidents occur, the business focuses on understanding what went wrong instead of finding someone to punish.  
  • Leadership support: Leaders need to define and promote the vision and the goal that the business is working toward, fostering a blameless culture and giving teams autonomy to work out how the goal is achieved.  

Transformation inevitably involves setbacks. You’ll encounter technical debt from new experiments as team members learn and may encounter team members who resist the changes because “that’s how we’ve always done it”.  These challenges are normal and expected. The key is maintaining focus on your shared goal of delivering value and using it as the filter for every decision you make together.  

The goal should be the kingmaker for every decision – does this help us achieve our goal of delivering value to our customers? 

Metrics Tell a Story 

To know whether DevOps is working, you need to measure what matters.  While it’s important to measure uptime and other technical metrics, these don’t give you an understanding of the end-to-end delivery process. 

DevOps success requires capturing metrics that describe how the entire system works together, from concept to production: 

Metric 

What it captures 

What story it tells 

Deployment Frequency 

How often deployments occur. 

This reveals how consistent the delivery pipelines are and can show any misalignments that need to be addressed. 

Lead time for changes 

How long it takes to implement a change 

This shows bottlenecks in the entire delivery lifecycle and provides insight into any blockers that need to be addressed.  

Change failure rate 

The number of deployments that cause problems 

This reveals the optimisation of your delivery pipelines from a quality perspective – are your tests validating deployments? Or are they just checkboxes?  

Mean time to recovery (MTTR) 

How quickly you can fix and recover from an issue 

This reveals issues with observability that make it harder to diagnose a system, or technical debt that’s creeping in and slowly making changes more difficult to recover from.  

These metrics work together to tell the story of your success.  If you deploy frequently with but a high failure rate, then you’re impacting customer experience; if you deploy reliable changes but very slowly, then you aren’t delivering value in a reasonable time frame.  

Tracking these metrics (and other metrics you find valuable) helps individual teams understand their impact toward the outcome.  They are usually supported by other metrics that are gathered from systems, such as uptime or failure-rate, or from delivery processes such as the rate of successful delivery pipeline executions, or time to resolve issues.  

Technical DevOps Outcomes that Drive Value 

DevOps culture enables teams to adopt new technical practices, to better operate systems, and to optimise the delivery model.  By implementing these practices, businesses can start to realise the value of DevOps.  

Observability: Get Useful Information 

You can’t improve what you can’t see.  Observability means having visibility into how your systems behave, how they’re used, and how they respond to real-world conditions.  Beyond log and metric collection, it’s about gathering data and making it meaningful enough to drive decision making.  

Modern observability often uses multiple tools, instrumented into applications and on servers to collect metrics, logs and traces, that often are enhanced with AI to spot patterns that could indicate potential issues or opportunities for optimisation.   

The data is often aggregated, enriched so that it can be understood, and presented through a dashboard or query tool, for administrators to view the state of the system and search through relevant data.   

When customers report slow load times, DevOps-enabled teams can look at dashboards or query logs and traces to gain an understanding of the bottleneck.  When there’s an outage, they can quickly identify the root cause.  

Reliability: Automated Operations 

Reliable systems behave how they’re supposed to, even when put under unexpected load or faced with component failures.  Achieving reliability isn’t just a case of adopting resilient architectural patterns; it’s about automating systems to respond to abnormal situations. 

Data gained from observability enhancements can be used to understand system behaviour and state. DevOps teams can use this data to make automated decisions as well, such as scaling out servers to accommodate increased load or pointing a DNS record at another server to recover from a site outage.  

Instead of bolstering your systems with more unnecessary infrastructure to handle outlier scenarios, by implementing automation to respond to those scenarios with the data you’re gathering, you can significantly enhance your system’s ability to recover when under stress.  

Speed: Time-to-Value 

Speed means getting new features to customers quickly without an increase in failure rates.  With the combination of automation to handle unexpected scenarios and lots of data, DevOps teams can start iterating faster and with higher confidence.  

  • Continuous Integration (CI) pipelines can be configured to scan code for security vulnerabilities, run tests to validate system functionality and performance, and use the metrics to ensure the code is safe to release.  
  • Automated version tagging, documentation, and release-note generation can speed up the development process and remove toil.  
  • Continuous Delivery (CD) pipelines can deploy and roll back releases, request approvals, and run smoke tests to validate deployments are successful.  

By wrapping the end-to-end delivery process in automation, releases are faster and safer.  It’s important to have an observable and reliable system designed or running first to ensure you have practices and processes to build your automation around.  

Efficiency: Cost-to-Value 

Efficiency is the pinnacle outcome of DevOps-enabled teams. It’s the combination of all other outcomes with constant improvement.  As teams build efficient systems, the systems will become more dynamic and re-usable, allowing you to quickly implement these practices on other systems.  

Some examples of efficient system practices are:  

  • Automated resource scheduling: Automatically scheduling and scaling resources to optimise performance and cost. 
  • AI-assisted troubleshooting: Combining existing observability data with LLM agents to automate initial troubleshooting and identify problematic releases.  
  • Ephemeral environments: Creating temporary non-production environments for testing on an as-needed basis, minimising infrastructure costs.  

Over time, DevOps becomes a runway for adopting new technologies quickly and meaningfully. 

DevOps as a Catalyst for AI Adoption 

The same cultural and technical principles that enable organisations to adopt DevOps are also used to create and operate scalable, production-ready AI applications.  

Without a good DevOps foundation, AI solutions become increasingly expensive and complex to manage.  They struggle to reach production-grade quality, resulting in either a vanilla solution that isn’t customised to the organisation’s specific needs or an over-engineered solution that collapses under its operational complexity.  

MLOps: DevOps for Machine Learning 

AI applications face unique challenges that traditional software doesn’t have.  AI models are probabilistic systems trained on data that can easily become stale, creating several problems:  

  • Data processing: Training and fine-tuning AI models requires massive amounts of high-quality data from multiple sources and in multiple formats.   
  • Model decay: AI models degrade over time as the data they are trained on becomes older and less relevant.  
  • Training infrastructure: Training requires significant compute resources that are expensive to maintain.  
  • Compliance requirements: AI decisions need to be explainable and auditable.  

MLOps applies DevOps principles to solve these challenges through automation, version control for data and models, and model observability.   

Data Operations: The Key to Implementing AI 

A model is only as good as its data, or the data it is using as context.  Managing data at scale requires the same disciplines as automating infrastructure and software delivery.  

  • Data pipelines: Automate the ingestion of data from multiple sources, transforming it into multiple formats and ensuring quality. Beyond ELT, to manage AI systems effectively, datasets should be versioned, deduplicated, tested, and validated.  
  • Scalable infrastructure: Provision and scale out resources for data processing workloads that can facilitate the processing of large amounts of data asynchronously and on a schedule.  
  • Governance and security: For legal and security reasons, data needs to be anonymised, and potentially harmful content removed. Often this process requires orchestration of multiple services to identify personal and/or harmful content in data so that it can be removed.  

Model Observability: Optimising your Machine Learning Applications 

AI applications fail differently than traditional software: They rarely crash, instead providing poor responses and hallucinations that reduce the benefits AI can bring to your business, making observability critical to a successful AI application. 

  • Performance monitoring: Tracking metrics like response latency, response accuracy, and response relevance can indicate when AI models require re-training or fine-tuning.  
  • Cost optimisation: Token usage and prompt tracing is vital to reduce the costs of AI applications and provide indicators of where to optimise your AI models or data.   
  • Auditability: Not logging or tracing your organisation’s model use can leave you vulnerable to security risks through model exploitation or misuse.  

Leadership Takeaway 

Innovation doesn’t scale without structure, and DevOps is the structure that makes speed possible. For leaders, this means: 

  1. Assessing your current DevOps maturity 
  2. Aligning your teams around shared values and outcomes 
  3. Investing in the right people, processes and partners 

How Inde Can Help 

DevOps is more than a set of tools or processes. It is a strategic capability that helps organisations innovate faster, operate more efficiently, and respond to change with confidence. At Inde, we help businesses realise the full potential of DevOps by aligning teams, technology, and strategy around shared outcomes. 

Whether you are laying the groundwork or looking to accelerate existing initiatives, Inde can support your DevOps journey. Contact us to learn more


Subscribe to our newsletter for more Inde insights

avatar
About the author

Michael Jarry

Michael's journey into tech wasn't exactly conventional. After completing the Ministry of Social Development's Limited-Service Volunteers (LSV) program, he started his IT career on the service desk, which led him into an operations team, branched into AWS DevOps consulting, and then landed at Inde, where he's become the resident AWS and DevOps advocate. Over his career, Michael has worked on projects ranging from automating enterprise private cloud deployments to developing platforms used by engineers to provision cloud infrastructure on AWS. Regardless of the project, the story remains the same—he’s focused on making lives simpler through DevOps and automation. As a long-time AWS and DevOps practitioner, Michael shares his knowledge through local AWS events and online.

COMMENTS