New AI releases tend to spark two reactions in IT: excitement about productivity and anxiety about displacement. A recent discussion around Anthropic’s new AI tool—amplified by public commentary from Zoho founder Sridhar Vembu—highlights a core reality: the next wave of AI isn’t only about better chatbots, but about tools that can perform larger slices of real IT work.

What’s different about “AI tools” vs. classic chatbots

ChatGPT-style interfaces popularized conversational assistance: ask a question, get an answer. The newer category of AI products is moving toward execution rather than just advice. In practical terms, this shift typically means:

  • Workflow integration: AI that connects to repos, ticket systems, documentation, logs, and internal knowledge bases.
  • Task completion: drafting code changes, generating tests, proposing fixes, summarizing incidents, or preparing change requests.
  • Multi-step reasoning: handling sequences like “read the error → inspect config → propose patch → explain impact → generate rollout plan.”

When AI tools become embedded in day-to-day systems, their impact on roles can be more direct than a standalone chat experience.

Why the IT jobs conversation is resurfacing

Warnings about job impact are not necessarily about all IT work disappearing. They are usually about a rebalancing of tasks:

  • Routine tasks shrink: repetitive scripting, boilerplate code, standard ticket responses, and basic troubleshooting become faster or partially automated.
  • Output expectations rise: the same team is expected to ship more, handle more systems, or shorten incident and delivery cycles.
  • Skill premiums shift: higher value moves to architecture, system design, domain expertise, security, reliability engineering, and careful oversight.

This is why leaders sometimes issue “warnings”: not because people have no future in IT, but because the composition of valuable work changes quickly.

What roles and tasks are most exposed

In many organizations, exposure correlates less with job titles and more with task profiles. Areas commonly affected first include:

  • L1/L2 support patterns: common “known issue” tickets, password/access workflows, and standard troubleshooting scripts.
  • CRUD-heavy development: repetitive endpoints, scaffolding, migrations, and templated UI components.
  • Documentation and reporting: meeting notes, postmortem drafts, release notes, and status updates.
  • Basic data work: query drafting, simple transformations, and summarizing dashboards.

That said, the same tools can also reduce toil for senior engineers, enabling more focus on design, quality, and risk management.

What doesn’t get “solved” by tools alone

Even highly capable AI tools tend to struggle—or require strong human governance—in areas such as:

  • Accountability and risk: who signs off when an AI-generated change causes an outage or compliance issue?
  • Context and priorities: understanding what matters to the business, what can wait, and what must be escalated.
  • Security and privacy constraints: safe handling of credentials, secrets, regulated data, and internal IP.
  • System-wide reasoning: emergent behavior in distributed systems, performance tradeoffs, and complex dependencies.

These gaps are often where careers can grow: turning into the person who can define constraints, evaluate outputs, and steer outcomes.

Practical guidance for IT professionals

  • Become “AI-supervised” rather than “AI-replaced”: learn to review AI output like you review junior engineers’ work—tests, threat modeling, edge cases, rollback plans.
  • Strengthen fundamentals: networking, OS, databases, security, and distributed systems knowledge becomes more valuable when implementation gets faster.
  • Own a domain: deep knowledge of an industry system (fintech, healthcare, ERP, logistics) is harder to automate than generic coding.
  • Measure impact: track lead time, incident rate, MTTR, and defect escape rate—teams that can prove outcomes get investment.

What leaders should do before “AI adoption” becomes chaos

If AI tools are introduced without governance, organizations often see inconsistent quality, security risks, and hard-to-debug changes. Useful steps include:

  • Define approved use cases: e.g., documentation and test generation first; production changes gated behind review.
  • Set policy for data access: what can be sent to external services, what must stay internal, and how logs/prompts are stored.
  • Update SDLC controls: code review rules, CI test requirements, SBOM/dependency checks, and change management.
  • Invest in enablement: training on prompting, evaluation, secure coding, and reliable incident workflows.

How to think about ChatGPT alternatives in this context

When comparing ChatGPT alternatives or new AI tools, the most important question is not “which model is smarter,” but:

  • Can it safely connect to your tools? (repos, tickets, docs, monitoring)
  • Can you control data boundaries?
  • Does it support review and audit?
  • Does it reduce toil without increasing risk?

In other words, the “best” alternative is the one that improves throughput while preserving reliability, security, and accountability.

Bottom line

Anthropic’s new tool and the reactions it triggered are a signal: IT work is shifting from manually executing every step to orchestrating, validating, and governing increasingly capable AI systems. The safest strategy—for individuals and companies alike—is to treat AI as a force multiplier and to specialize in the parts of IT that require judgment, ownership, and systems thinking.