The state of AI with Aditya Shrivastava, Wise’s AI & Automation Specialist

Madi Corr

We sat down with Aditya Shrivastava, AI & Automation Specialist at Wise, to answer three frequently asked questions on the state of AI this month.

How has the corporate perspective on AI evolved over the past year? Are executives thinking about the technology differently today than they were last summer?

The executive perspective on AI has matured rapidly over the last 12 months. We’ve transitioned from curiosity to strict operational discipline.

Today, AI is understood as a strategic moat, not a toy. Leadership has realized that the technology is out there and available to everyone — the differentiator is the workforce. If your team isn’t actively skilled in using these tools, your competitors are widening the gap.

Simultaneously, the era of go-wild experimentation is ending as the reality of costs are setting in. Teams are increasingly being asked to justify the business case, estimate ROI, and prove cycle-time reductions before any workflow is greenlit. Plus, liability is personal: "The AI made a mistake" is no longer an acceptable excuse to clients or regulators. Companies are enforcing strict "human-in-the-loop" guardrails, holding teams directly accountable for the accuracy and security of whatever AI outputs.

Many businesses are realising the financial consequences of maximising their AI token consumption. How can businesses lower their spend without giving up the tools that help power their productivity?

It’s important for teams to understand ‘context hygiene’. Every time you send a message in a long, continuous thread, the system re-reads the entire chat history. The solution is to stop letting single chat windows run for days. Start a clean, new chat session every time you switch tasks. This simple behavioral shift can cut your token consumption exponentially.

Another solution is to stop chasing shiny-new AI tools. In the early rush, many businesses bought expensive software subscriptions for tools that are wrappers built on top of public APIs. So it's worth evaluating where you can build your own light tools. Instead of paying $50 per seat per month to a middleman, have your developers build simple internal interfaces that connect directly to raw APIs. You will only pay for what you actually use, bypassing the SaaS markup entirely.

For the vast majority of daily business tasks - like sorting spreadsheets, basic coding, or summarizing meetings - you do not need the latest and greatest flagship model at an "extra high" thinking setting. A fast, low-cost "mini" model will do the job perfectly for a fraction of the cost.

A simple way to get started is to:

  1. Run your team’s specific workflow through the absolute best, most expensive model available. If the output is excellent, that is your baseline.
  2. And then start stepping down systematically: run the exact same prompt/workflow through the next tier down.
  3. Find the cliff: Keep downgrading until the output quality finally starts to suffer. The model immediately before the quality dropped is your sweet spot.

How should businesses decide which AI tool makes the most sense for their team? What criteria should they consider?

In addition to the cost aspects mentioned above, run every tool through these filters:

  • Does it integrate naturally into where your team already works (e.g., Slack, Confluence, Google Workspace)? If it requires copy-pasting between tabs, your team won't use it.

  • Does the vendor have a ‘Zero Data Retention’ policy, or are they using your proprietary company data to train their future public models?

  • Does the tool support modern, open connection standards like MCP (Model Context Protocol)? If it doesn't "talk" to your existing databases, it’s a silo.

These criteria should help you to understand how valuable a tool will be for your specific use case.

Ready for more more AI insights?

Read our blog


*Please see terms of use and product availability for your region or visit Wise fees and pricing for the most up to date pricing and fee information.

This publication is provided for general information purposes and does not constitute legal, tax or other professional advice from Wise Payments Limited or its subsidiaries and its affiliates, and it is not intended as a substitute for obtaining advice from a financial advisor or any other professional.

We make no representations, warranties or guarantees, whether expressed or implied, that the content in the publication is accurate, complete or up to date.

Money without borders

Find out more

Tips, news and updates for your location