### Cutting the Cost of AI - A Guide to Token Efficiency

**Most organisations run AI or AI agents in production without a clear picture of what it actually costs. Every request consumes tokens both as input and output, and those tokens translate directly into spend that's often underestimated. Because hands-on experience with AI remains limited in many companies, these costs remain unclear and, particularly for agents, can add up to substantial sums before anyone notices.**

In this live session, we show how organisations can make their AI costs visible and bring them under control. Using a live comparison, we demonstrate just how much prompt design and model choice affect token usage and cost. You'll see what to look for when identifying cost drivers, and how this kind of tracking can be set up to run continuously, depending on your organisation’s setup.

Experience live demonstrations on tracking token consumption, measuring costs, and identifying effective ways to reduce expenses while maintaining high standards. Organisations frequently deploy AI in production without a true grasp of the actual costs involved. Every request consumes tokens; each model is priced differently, and with AI agents in play, expenses can escalate rapidly and go unnoticed. 

What if, in one hour, you could see live just how big these differences really are and which levers actually move the needle?

### Agenda

* 5 min — The problem. Why token costs stay invisible inside most organisations, and what changes once agents are in the loop.

* 30–40 min — The cost breakdown.

  * Results across different prompt styles and models, shown as a visual comparison

  * A live agent optimisation — same task, different cost, same output quality.

  * What "token maxing" is and why it's one of the largest controllable cost drivers we see.

* 10–15 min — Q&A and what's next. Your questions on the approach and the method

### Who it's for

* Executives accountable for AI spend who want predictability rather than surprises.

* Teams running or rolling out AI agents who suspect costs are higher than they should be, but lack visibility into why.

* Anyone who currently experiences AI costs as a black box.

### What you'll take away

* A concrete sense of how much prompt wording and model choice can move your costs — backed by real comparison data.

* A clear understanding of "token maxing" and how to spot it in your own organisation.

* A realistic view of where agent workflows put your budget most at risk.

* What a deeper, tailored analysis of your own usage and workflows could look like — same method, your data.

### Practical

- **Duration:** 1 hour
- **Format:** Virtual (Zoom)
- **Cost:** free
- **What to bring:** any questions and experience you have regarding cutting AI token cost


### Your Speaker

**Leona Schulz** – Hitabis GmbH

Leona Schulz is a full-stack software developer at Hitabis and specializes in the development of AI agents and the efficient use of AI.

Among other things, she has contributed to a GreenIT project and brings practical experience in how AI technologies can be integrated into software products in a meaningful and cost-effective way.

[Hitabis GmbH](https://www.hitabis.de), a Berlin-based IT services company, has been developing customized software solutions for businesses, government agencies, and organizations since 2004. Its range of services extends from the development and integration of technical systems and machinery, through database and web development, to the successful implementation of complex digitization and migration projects.

In this live session, we show how organisations can make their AI costs visible and bring them under control. 

Cutting the Cost of AI - A Guide to Token Efficiency

### Re-Think Your Operating Model — Live, with AI

**A 1-hour interactive session for Operating Model Owners, Transformation Owners, and modelling practitioners. We start with an empty canvas and build a knowledge graph of an operating model from scratch — live, with you, with AI in the loop.**

Most enterprises run on operating models that live in tacit knowledge, scattered documents, and the heads of senior people.

What if you could turn that mess into a queryable knowledge graph — in real time, with AI doing the heavy lifting?

In this 1-hour live session you'll watch (and join in on) the build itself. We put a fictitious company on the table, share the kind of artefacts most enterprises actually have — a leadership memo, a process write-up, a KPI snapshot, a strategy slide or two — and build a knowledge graph live with AI. 

No prepared demo, no polished result. Just the workflow, as it happens.

We:
  
  - Describe the situation to the AI assistant.
  - Share the documents and watch the assistant parse them.
  - Build the knowledge graph node by node — capabilities, processes, decision rights, KPIs, dependencies — with the audience suggesting what to ask next.
  - Surface tensions, gaps, and the decisions that fall out.


### Agenda

- **5 min — The problem.** Why operating-model conversations stall on slide decks and out-of-date documents — and what changes when you can build the model collaboratively, with AI, in the room.
- **30–40 min — The live build.** A fictitious company (we'll share the brief on the day). A handful of typical artefacts. An empty metapad model.
- **10–15 min — Q&A and what's next.** Your questions on the build, the platform, the method; a walk-through of our **AI-Driven Operating Model Workshop** (our full-day version on *your* operating model, with your real documents)


### Who it's for

- **Operating Model Owners** — CEOs, COOs, CTOs and other senior leaders accountable for how the enterprise actually runs, who want their operating model to be a queryable, improvable artefact rather than tacit knowledge spread across people and slide decks.
- **Transformation Owners** — accountable for a specific multi-stakeholder enterprise transformation, looking for ways to keep the model of the work explicit and useful as the programme evolves.
- **Modelling practitioners** — enterprise architects, business engineers, knowledge engineers and operations researchers who already work with structural and process models and want to see what an AI-grounded, computational version of that practice looks like.
- **Anyone curious** about how AI assistants behave differently when grounded in a typed knowledge graph of an enterprise rather than scraping a pile of documents.

### What you'll take away

- A first-hand view of the *workflow*: how a knowledge graph gets built from prose-and-doc inputs, live, with AI.
- A clear picture of the difference between AI-on-documents (RAG) and AI-as-co-modeller (a *computational knowledge graph* being built in front of you).
- A concrete next step if you want to do this on your own operating model — same workflow, your data, more time.

### Practical

- **Duration:** 1 hour
- **Format:** Virtual (Zoom)
- **Cost:** free
- **What to bring:** questions you'd ask your own operating model if it could answer; a willingness to suggest what to ask next during the live build


## Your Host

**Dr. Oliver Grasl** - Founder & Managing Partner at transentis

With over 25 years in enterprise transformation, Oliver brings deep expertise in business engineering, operations research, and analytics to reshaping enterprises operating models.

Oliver's passion is building digital twins that bring human intelligence and machine intelligence together. 

He believes the best decisions emerge from this collaboration — and builds "computational knowledge graphs" that make this possible, turning complexity into clarity and making change steerable.

Live demo: See how we build a live, queryable knowledge graph of an enterprises operating model

Re-Think Your Operating Model with AI

How customer service shifts from an agent handling every ticket to AI handling the routine ones with a person on call, and where that quietly goes wrong.

How AI Changes the Issue-to-Resolution Value Stream

Services

Products

Training

Resources

Company

Chrissoula Sourli

Finances & Office Management

Know-how

Market Experience

Contact

transentis consulting

Geisbergstraße 9

10777 Berlin