The right AI for every step. Lower costs per result.

AITJE It maps out how your organisation uses AI and what that costs. It then examines subscriptions, model selection, agents and workflows, and makes adjustments where agreed. The aim is to reduce unnecessary usage whilst maintaining the required quality. Read more about Token management & optimisation.

Both. The AI scan provides a broad overview of your current AI usage, subscriptions and opportunities for optimisation. A comprehensive model comparison, workflow adjustments and ongoing management are covered under a separately agreed engagement. You can also purchase this service directly, without first carrying out a scan.

No. ‘AITJE’ also considers AI usage via subscriptions and applications for content, knowledge, images, code and other tasks. Not every service is billed in tokens: for images, audio or video, for example, credits, the number of images or minutes may be relevant. This usage data is factored into the cost per usable result.

That’s possible. Multiple smaller subscriptions, a shared model budget or API usage may be a better fit than the current setup. AITJE compares actual usage, included capacity, limits, quality and terms and conditions. Which combination is right depends on your team and tasks.

Groups of agents, each carrying out a specific part of the work, such as retrieving information, writing text and checking the result. Each step can be assigned a suitable model and specific context. AITJE assesses whether this division of labour adds value to your workflow; additional agents may also mean extra calls and management overhead.

Yes. AITJE can combine advanced models via external APIs with models on its own servers or local devices. Edge models run on, or close to, the device where the task takes place. The task, capacity and data flows determine the most suitable approach. With in-house models, hardware, power, hosting and management remain part of the costs.

No. ‘AITJE’ tests your tasks and compares quality, speed, usage and the number of attempts. A model with a lower token price may require more calls; a more expensive model may actually work out cheaper for a particular step.

This is a sample calculation, not a firm commitment. Assuming a hypothetical flat rate of €4 per million tokens, one million tokens would cost €4 and one hundred thousand tokens would cost €0.40. The achievable savings depend on the workflow, output, quality and retries. Costs for code, hosting, management and development are separate from this.

For example, retrieving data via an API, selecting relevant fields, performing calculations and checking the output against fixed rules. The model is given a clearly defined task with specific context, rather than all the raw data and steps in a single large prompt.

Yes. AITJE can build a central layer for model calls and a new workflow, including routing, logging, budgets and limits. Access, key management and a form of ongoing management are also agreed in advance.

This service focuses on AI usage and costs within your organisation: subscriptions, models, agents and workflows. Optimisation assesses an existing AI solution more broadly, for example in terms of reliability, stability and human oversight.

By testing the end result. ‘AITJE’ compares the original and the adapted workflow on the same tasks and assesses whether the outcome still meets the agreed requirements. Fewer tokens or a cheaper model only constitutes an improvement if the result remains good enough.

Yes, provided that usage is measurable and can be attributed. AITJE monitors subscriptions, model calls, token usage and other consumption per task or team. Where necessary, logging can be set up to gain that insight.

AITJE A fallback can be built: an alternative route or a different model that takes over the task. For critical steps, a local or edge model may be a good choice. Which fallback is possible depends on the task and is agreed upon during the setup.

AITJE It can incorporate limits on attempts and costs, logging and clear stopping points. At key moments, an employee may first need to assess whether the workflow can continue: ‘human in the loop’. In this way, technical limits are combined with human oversight.