What We Think

There’s no denying that artificial intelligence is the belle of the ball, well, the belle of ALL the balls. Apparently, it is going to make your (insert buzzword here) a world-class growth engine that will (insert another buzzword-heavy phrase here).

But there is a dirty little secret that affects every organization – particularly large enterprises. AI doesn’t know the quality of your data.

But there is a dirty little secret that affects every organization – particularly large enterprises. AI doesn’t know the quality of your data. It will loudly proclaim a convincing, authoritative narrative, even if the underlying data it is based on is absolute trash. And let’s be honest here – most of the time it is.

For most organizations, defining KPIs and aligning customer touchpoints with the data collection to support them is an afterthought – if it’s even a thought at all. This is a significant problem in a single system, but modern digital ecosystems are a patchwork of systems that need to be integrated and normalized, snowballing the issue. This is nothing new, of course. Most senior leadership don’t understand analytics or data science. They view it as just a cost center that should be as lean as possible, rather than the key to defining success and understanding which efforts are most effective at executing strategy and driving revenue.

For decades, technologists have acknowledged the phrase ‘Garbage in, garbage out’. But in today’s frenzied AI FOMO environment, it’s being ignored. Unfortunately, this is incredibly dangerous. It can lead to basing an entire digital strategy on a confident, persuasive argument that is no better than a hallucination. Layering on agentic AI makes this infinitely worse.

In the old days, broken data would reveal itself in a wonky dashboard. An analyst would dig into the issue and identify it. AI doesn’t work that way. It’s designed for ambiguity and to fill in the gaps. In other words, AI isn’t going to tell you it’s wrong. It will invent a reason why it’s correct. The combination of both agentic AI and leadership making autonomous decisions based on a flawed picture of the landscape is inherently hazardous. This synthetic authority is a critical risk to an organization. It’s not only a challenge to big swings – smaller, frequent decisions are arguably even more of a problem as they slowly compound distortions of reality until its entirety is flawed.

Does this mean organizations shouldn’t adopt AI? Of course not. It means that AI should not be operationalized before an organization’s data house is in order and prepared for the roadmap ahead.

But what does that mean? It means a strong data strategy with solid governance. Data is not just a byproduct of operations. It’s a product and an asset whose management, curation, and documentation should be a top organizational priority.

Most of us in the technology space are familiar with the old project management adage that you can only pick 2 options: Good, Fast, or Cheap. Up until now, that was ironclad. For the first time, that tradeoff is becoming negotiable, but perhaps not in the way one would imagine.

Countless leaders envision AI dancing through their development ecosystem, magically creating new products with little need for the challenges that come with human designers or developers. Currently, that is a dangerous pipe dream. AI will change over time, but organizations that do not maintain oversight by skilled individuals will spend far more than they save to fix what they create.

The organizations that gain the most from AI over the next several years will not be the ones chasing fully autonomous software development. They will be the ones applying AI surgically to high-friction operational bottlenecks with measurable ROI.

But there is an area where AI is a paradigm shift that will enable success today. Ironically, one of AI’s clearest near-term enterprise use cases may sit in one of the least glamorous areas of software delivery: QA. Yes, that team that is always giving developers defects they cannot reproduce, or that finds a critical defect right before the team is supposed to ship.

Automated testing is not new. Teams have been rolling it into their release pipelines for quite a while now. However, while it helps speed up both testing for new features and regression testing, it has a few major drawbacks: creating and maintaining automated tests is both expensive and time-consuming; changes to functionality or UI can easily break the tests; and scripts and test data need to be recreated repeatedly.

As a result, to make automated testing a viable advantage, an organization requires a sizeable, dedicated team of automation engineers not only to create the initial tests but also to continually hunt down issues that break tests and fix them. It’s an expensive proposition.

Not anymore. How do we know this? Easy. We built it. Imagine this scenario: Mid-sprint, the QA team is ready to test a new feature. With the press of a button, our platform reviews the user stories, and within minutes, not days, a full suite of test scripts is created, including edge cases. Days later, a developer changes a button within the feature. Previously, that would have been a broken test that would have taken half a day of detective work to fix. With our platform, however, the test heals itself and keeps running. In case of a serious failure, a screenshot and video show exactly what broke and where it is. Triage has been reduced from hours to minutes, and the team ships on time.

For the CTO, that means faster, more predictable releases with less firefighting. The QA manager sees their team focused on actual high-priority issues, rather than manually creating scripts and chasing ghosts. Product and Marketing have an effective, brand-strengthening outcome, and Finance recognizes the value of a small automation team doing the work of a large one.

The organizations that gain the most from AI over the next several years will not be the ones chasing fully autonomous software development. They will be the ones applying AI surgically to high-friction operational bottlenecks with measurable ROI. QA is one of the clearest examples available today. That is exactly why we built TestWyze.

Ready to stop choosing 2 options? TestWyze helps organizations modernize QA with AI-driven automation that scales with the pace of development. Let it help you get Fast, Cheap, and Good.

We get it. Every company on the planet is asking itself and its partners what it can do with Artificial Intelligence. We get asked the question constantly. No one likes the answer: “Well, it depends.” Unfortunately, it’s correct.

Organizations cannot keep spending on AI like this without expecting a revolt, at best. The question is whether they are going to revisit their strategy now or hope that AI gets much better soon enough so the waste won’t matter.

Oxford Economics estimates the US AI enterprise spend was roughly $200 billion in 2025. US companies spent $1 trillion on all R&D in 2025, according to the National Science Foundation. That means AI accounted for 20% of ALL R&D spending for 2025, and it’s increasing rapidly.

That’s a huge number. Obviously, the next question from anyone fiscally inclined is “What value did we derive from that massive spend?” That’s when the consultants look back at them with a blank stare. The ROI on that spend is MISERABLE. Deloitte reports that only 6% of investments had a payback in under a year, and even among the best efforts, only 13% saw returns in that same period. In fact, AI pilot failure rate estimates range from 70% to 95%. By any measure, that’s insane.

Einstein never said, “The definition of insanity is doing the same thing over and over and expecting different results,” but that doesn’t make it any less true. Organizations cannot keep spending on AI like this without expecting a revolt, at best. The question is whether they are going to revisit their strategy now or hope that AI gets much better soon enough so the waste won’t matter.

We are definitely in camp #1. Don’t get us wrong, we are not advocating that AI spending stop. Not at all. We firmly believe that AI will add tremendous value in the long run. However, we do believe the throwing of AI spaghetti at the wall needs to end. Choosing AI use cases based on enthusiasm or fear that you need to do something because your competitors are is not a recipe for favorable business outcomes. It will only make future efforts harder.

AI adoption will not be easy. There are many variables that factor into whether an AI effort will succeed or fail. And even with favorable variables, AI is frequently and confidently wrong. A lot. That’s why a far more deliberate strategy is needed.

Our advice to organizations that are frustrated with a lack of AI success and those who are paralyzed by the breadth of the choices is the same: Take your cold shower. Learn to prioritize where to focus your resources. Used correctly, AI can be a strong accelerator for making your business better at what it does well. It cannot fix areas that already don’t work because they are immature.

Figuring out the right workflows for AI adoption does not need to be an exhaustive analysis of every single flow, system, and business unit in the organization. Talking to the right people will quickly reveal patterns across the organization that will allow for a solid 18-month roadmap that lets you focus on what will work now and what should be planned for.

Not sure where to start? We can help navigate through the noise and find your footing. Our AI Reality Check will give you the intelligence and confidence to move forward.

Ah, marketing pixels, those simple little JavaScript snippets that your marketing team gives you the night before a new campaign is supposed to launch. The ones that always work perfectly out of the box, never require customization, and always come with a clearly defined retirement date. Exactly those.

Not having tight control over those tiny little scripts in your tag management system can cause huge headaches, including privacy and regulatory compliance liability, broken conversion tracking, site performance issues, downstream retargeting failures, and security risks.

When was the last time you thought about them? How many of them are sitting in your tag management system right now? How many of them are broken because the event they depended on was changed months ago? Or worse, how many of them are still silently sending customer information to a third party with whom you no longer have an agreement in force?

Admit it, you’re just like almost all the marketing technology teams out there, you have no idea. Does your CISO know this? How about Legal? We understand. Analytics and MarTech are rarely in the forefront of teams’ planning. You love the idea of governance, but your team is always spending so much time dealing with last-minute requests like these, making sure new functionality is covered, and running ad hoc reporting analyses that it never happens.

But it matters. A lot. Not having tight control over those tiny little scripts in your tag management system can cause huge headaches, including privacy and regulatory compliance liability, broken conversion tracking, site performance issues, downstream retargeting failures, and security risks.

These tags often collect customer and behavioral data, whether they’re being actively provided or not. Lack of governance can expose organizations to violations of constantly evolving regulations, including HIPAA, CCPA, GDPR, and numerous others that are enacted each year. Not knowing what is being collected and where it’s going can, at best, lead to a fire drill when your organization is accused of infringing on this legislation, and, at worst, lead to actual penalties.

Most major e-commerce organizations rely heavily on third-party advertising and marketplace platforms to drive revenue. Broken conversion tracking causes a nightmare to reconcile, which erodes trust with both the retailer and the platform, and makes marketing spend attribution that much harder to determine.

It’s well documented that slower page load times negatively affect site conversions and SEO rankings. Industry studies show that a single second delay can reduce conversions by up to 20%. Too many, poorly configured, or broken tags can demonstrably cause slower page load times, leaving your team a prime scapegoat for the sales team’s failure to meet their quarterly goals.

Finally, these tags are designed to allow 3^rd parties to run code on your site without involving your engineering teams. This means they can download other scripts or perform other functions you may be completely unaware of. Your legal department will love hearing this reinforced when they ask for a list of everything that is being served on the site and what they are doing.

A good governance system alleviates these risks by making sure organizations have a registry that, at a minimum, tracks:

• Tag inventory and documentation
• What is being tracked
• How long it is tracked for
• Where the data goes
• Data layer standards
• Quality engineering procedures

But even a good governance system is only effective if it’s accurate. It’s also critical to periodically audit your tag management ecosystem to ensure that these procedures are being followed. Not sure where to start? It’s understandable. Looking at an overgrown garden is always daunting. We can help you take the first step by conducting an audit of your current state and creating a good governance program to keep you running smoothly. We can even handle your entire tag workflow, so your team can focus on other priorities.