The AI Productivity Paradox in Sales Engineering

Artificial intelligence is advancing at an extraordinary pace. And yet the productivity boost we were all promised does not seem to be fully there.

From developers claiming their roles have fundamentally changed, to companies struggling to adopt AI at scale, there appears to be a gap between technological capability and measurable outcomes.

This article explores that gap through the lens of my own profession: cloud sales engineering. By breaking the role down into tasks and evaluating where AI helps most, we can better understand where productivity gains already exist and why they are not always visible at the organizational level.

As cloud sales engineers, we operate close to both the technology and the revenue it ultimately drives. This position puts us near the front line of AI adoption. The way our daily workflows have changed may therefore offer a clearer explanation of what is happening across the industry.

Exposure to Artificial Intelligence

A growing amount of research is trying to measure how AI will affect the workforce. Microsoft, for example, published the paper Measuring the Applicability of Generative AI to Occupations, while the International Labour Organization released a Refined Global Index of Occupational Exposure to Generative AI.

Most of these studies share a useful methodology: analyzing tasks within jobs rather than jobs themselves.

Instead of asking whether a profession disappears, the question becomes simpler and more practical: which parts of the work will be impacted and in what way.

The approach is straightforward:

1. Decompose the role into tasks

2. Evaluate AI capabilities

3. Assess the impact

Earlier research provides useful context. The 2013 paper The Future of Employment (Frey & Osborne) explored how automation might affect occupations long before the generative AI wave. More recent work attempts to refine those predictions with updated datasets and models.

Still, depending on assumptions, research often paints contradictory pictures. More importantly, many frameworks struggle to capture the lived reality of engineers working with AI today. That is why examining our own workflows can be equally valuable.

Sales engineering in the cloud era

Sales engineering appears in companies selling complex solutions, often in B2B environments. Simple products rarely require explanation, but modern digital platforms, cloud services, and security technologies are harder to adopt without technical guidance.

Coca-Cola does not require architecture diagrams. A distributed cloud platform usually does.

Sales engineers operate at the intersection of technical expertise and revenue generation. While sales representatives focus on commercial discussions, sales engineers help customers understand the technology, design solutions, and successfully implement them.

In cloud environments this role becomes even more important. Cloud services follow a subscription model, meaning success is not defined only by the initial sale. Customers must continue to derive value from the platform.

Convincing the customer is therefore only the beginning. Long-term success depends on ensuring the technology actually works for them.

In practice, the role revolves around three main activities:

Customer engagement. Sales engineers often act as the main technical contact for customers, running discovery workshops, discussing architecture, and sometimes helping when things go wrong.

Technical advisory. Engineers design solutions, demonstrate product capabilities, and frequently assist customers in building proof-of-concept environments.

Advocacy. Knowledge is scaled through presentations, technical content, and reusable assets that help the broader sales organization.

Within these areas, three tasks consistently have the highest impact in day-to-day work:

• running discovery workshops

• demonstrating proof of value through demos or proof of concepts

• helping customers overcome technical challenges

Before examining each task individually, it helps to define how AI typically affects work.

Across most research and industry discussions, AI impact generally falls into three categories: automation, augmentation, and transformation.

Automation replaces repetitive tasks.

Augmentation improves productivity while humans remain responsible.

Transformation changes how the task itself is performed.

Using this framework, we can estimate how AI affects the core activities of a cloud sales engineer.

Running discovery workshops

Discovery workshops are often the first deep interaction between a customer and the vendor’s technical team. The objective is not only to identify technical challenges but also to establish credibility and trust.

AI already helps significantly with preparation. It can summarize previous interactions, analyze the customer’s industry, generate discovery questions, and highlight potential architecture patterns.

During and after the meeting, AI tools can capture notes, summarize decisions, and generate follow-up actions.

But the interaction itself remains human. Discovery workshops involve navigating organizational dynamics, interpreting ambiguous answers, and understanding constraints that may never be written down. Stakeholders often have conflicting priorities or hidden limitations that only emerge during conversation.

AI can assist preparation and analysis, but the trust-building element remains largely human-driven.

Demonstrating proof of value

In cloud environments, value is rarely communicated through slides alone. Customers want to see working systems.

Demos and proof-of-concept environments allow them to validate assumptions and understand how a solution might behave in their own environment. Preparing these environments often involves repetitive setup work: generating datasets, building infrastructure, writing scripts, and documenting architecture.

AI can significantly accelerate these tasks. It can generate infrastructure templates, example datasets, demo scenarios, and architecture diagrams.

Engineers still need to validate designs and align them with customer requirements, but the experimentation cycle becomes dramatically faster.

Solving technical challenges

Eventually every customer encounters technical obstacles. Deployments fail, architectures do not scale as expected, or security policies block otherwise straightforward solutions.

This is where the sales engineer becomes a problem solver.

AI tools can assist by analyzing logs, searching documentation, proposing troubleshooting steps, and identifying configuration patterns. For well-documented issues, the time required to locate solutions drops significantly.

However, production environments are rarely simple. Problems usually involve multiple systems and constraints specific to the customer’s architecture. Engineers still need to validate suggestions and adapt solutions to real-world conditions.

The main change is speed. AI reduces the time required to explore possible solutions, but the final judgment remains human.

AI impact assessment on key tasks

For sales engineering, the pattern is clear: AI augments far more than it replaces.

A real-life example

To make this discussion less theoretical, consider a recent engagement involving a customer exploring a migration from their on-premises infrastructure to Oracle Cloud Infrastructure.

Their environment was complex and required several iterations of meetings, emails, and technical data collection. We needed exports of virtualized environments, operating system versions, hardware sizing, and software dependencies.

During preparation for the initial call, AI did not add much value. The focus was mostly on gathering information.

Things changed after the first meeting.

The customer mentioned technologies I was less familiar with, such as MonetDB, and specific limitations involving Elasticsearch features and PostgreSQL plugins. Research that would normally take days was reduced to hours. Instead of scheduling the follow-up meeting for the next week, we were able to meet again the following day.

Tools such as Cline together with models like Codex and Opus allowed me to spin up working environments quickly and validate migration strategies. For example, I experimented with a migration from Elasticsearch to OpenSearch while dealing with version constraints. With AI assistance, I was able to prototype and test the approach in less than a working day.

Preparing the final presentation meant consolidating everything: documenting findings, preparing a tailored demo, and ensuring a working environment was ready to showcase the migration path.

The engagement itself lasted two to three weeks due to asynchronous communication with the customer. But if I look strictly at focused engineering time, I estimate the work required no more than sixteen hours.

My rough productivity estimates were the following:

This suggests that AI allowed me to compress roughly a week and a half of work into two focused working days.

Where the gap comes from

If productivity gains at the individual level are already visible, why do we not see the same acceleration at the organizational level?

The answer may lie in a simple concept borrowed from physics: inertia. Moving a stationary object requires significantly more energy than keeping it moving. Anyone who has tried pushing a car knows this feeling. The first few centimeters require the most effort. Once the car starts rolling, however, each additional push becomes easier and the movement accelerates.

Technology adoption often behaves in a similar way.

Even if one engineer becomes significantly more productive, the broader system still moves at its previous speed. Customers still need time to gather information, align stakeholders, schedule meetings, and validate decisions. Procurement processes, architectural reviews, and risk assessments operate at their own pace.

Economists have observed similar dynamics before. Robert Solow famously noted in 1987 what became known as the productivity paradox:

“You can see the computer age everywhere but in the productivity statistics.”

Innovation research describes the same phenomenon through Everett Rogers’ Diffusion of Innovations model. New technologies typically spread slowly at first, while early adopters experiment and validate their usefulness. Only after confidence builds does adoption accelerate across the broader market.

From this perspective, the productivity gains we observe today may represent potential energy building up within the system. Faster research, quicker prototypes, and clearer demonstrations gradually reduce uncertainty for customers.

At some point the system starts moving faster.

Closing thoughts

Looking at these tasks individually reveals a clear pattern. AI does not eliminate the need for cloud sales engineers. Instead, it changes how we work.

Much of the repetitive preparation, research, and documentation can now be automated or dramatically accelerated. Engineers can prototype architectures faster, validate solutions earlier, and deliver higher-quality outputs in less time.

At the task level, productivity gains are already real.

Yet organizations and customers move at a different speed. Until their processes adapt, much of the acceleration remains partially hidden.

This is the essence of the AI productivity paradox.

The technology is already changing how engineers work. The real challenge is whether organizations can adapt quickly enough to capture the momentum it creates. In systems theory this is well understood: optimizing a component does not optimize the system.

And much like pushing a car, the hardest part may simply be getting the system moving.

References

Solow, R. (1987). We’d Better Watch Out. New York Times Book Review.

Frey, C. B., & Osborne, M. (2013). The Future of Employment: How Susceptible Are Jobs to Computerisation?

Rogers, E. (2003). Diffusion of Innovations.

Microsoft Research. Measuring the Applicability of Generative AI to Occupations.

International Labour Organization. Generative AI and Jobs: A Refined Global Index of Occupational Exposure.