How CAai works and what each of its functions does.
CAai supports you at every step of the storyboard methodology. Here you can read what each function does, what makes its result better and where the typical pitfalls lie. The structure follows the three phases of the work process.
Project – Define Basics
This phase establishes the foundation everything else depends on: who is speaking to whom, with what goal, on what factual basis. Almost every later function draws on what is defined here.
Create a new presentation
Your project description becomes a complete first draft in a single pass: briefing, intro, core message, storyline structure, chapters and arguments. You can then revise each part individually.
This improves the result: The project description is the single most important input in the whole application. Describe what it is about, who is listening, what you want to achieve and what speaks against it – rather one paragraph too many than one keyword too few.
Good to know: An unchanged run returns the same result for a few days, even if something else was adjusted in the meantime. To see a different version, change the description.
Revise presentation
You upload an existing presentation, and the application reads out its texts, images, charts and tables, assigns them to an argumentation structure and rebuilds it.
This improves the result: With "Existing content only", the chapters stay bound to what your file actually contains. Without this setting, the application fills in missing thoughts.
Good to know: With large files, the texts appear first and the images and charts only afterwards. That is not an error – the assignment waits until all content blocks are available.
Analysis
Assesses your project across several modules at once – briefing, sources, intro, core message, storyline, arguments and their coverage – and returns a value and concrete improvement suggestions for each area.
This improves the result: The more complete the briefing and the more solid the sources under "Materials", the more reliable the assessment.
Good to know: The values stay frozen until you run the analysis again. Accepting a suggestion therefore does not raise the value immediately.
Accelerator (Key Questions)
Asks structured key questions that help you describe your undertaking systematically. The answers move automatically into the matching briefing fields. "Fill in all fields" generates suggestions for everything still empty.
This improves the result: Answer in full sentences rather than keywords – the fields are derived from them, and little can be derived from a keyword.
Guidance
A conversation partner who guides you through the build-up with questions instead of handing you finished text. It knows your entire project state and uncovers blind spots.
This improves the result: It follows your order of work: you can change topic at any time. Where you disagree, it responds to that.
Good to know: The progress indicator measures how much has been clarified – not how good it is. And the guide knows your project, not your organisation; its estimates of time and scope are rules of thumb.
Briefing assessments
For individual briefing fields – such as challenges or difficulty – the application proposes a value together with a rationale, taking into account everything you have already filled in.
This improves the result: The more briefing fields are filled in, the more accurate the assessment. If a value is already in the field, it is read along and either confirmed or corrected with reasons.
Good to know: The difficulty is the most consequential entry in the entire briefing: it helps determine the form of argumentation and whether sensitive messages are offered as rhetorical questions. Review it rather than simply accepting it.
Case analysis
Assesses your starting position from the constellation of sender role, recipient role and business case: which topics carry weight, where the tensions lie, what this audience pays attention to. It is the basis of the argument suggestions.
This improves the result: Roles and industry are the axes of this analysis – the same role weights things completely differently depending on the industry. Without role information the result stays thin.
Good to know: The result is cached. If you change roles or the case, run the analysis again.
Reading sources: images, charts and figures
From uploaded files the application retrieves not only text but also images, charts and tables – including the values visible in them – and provides them as content blocks.
This improves the result: A PDF with a real text layer is the most accurate route. Labelled values in the original make the difference between an exact figure and an estimated one.
Good to know: Values recovered from a chart are a reconstruction, not a measurement – an "approx." in front of them is an honest statement, not an oversight. And a symbol or icon is deliberately not treated as content.
Relevance filter when reading in
With "Only thematically relevant content", the application takes from a file only what fits your project – instead of turning every marginal note into a content block.
Good to know: The yardstick is your project description. If it is vague, the filter works poorly. That no text content blocks are created at all can be the correct result.
Questions to your sources
You ask a question about your materials and receive an answer with references to where it was found. The search works by meaning rather than keywords – and across language boundaries, so an English document can answer a German question.
This improves the result: You choose whether only your sources may answer or whether general knowledge may be added.
Good to know: The search finds similarity, not truth. No hit does not mean that it is not in your documents – only that it is worded differently.
Assessment and recommendations
The analysis hub shows strengths, gaps and possible better wordings for each area – each with a confidence level indicating how solid the suggestion is.
Good to know: A suggestion in the hub can differ from what appears in the story area: they come from two different generators. What counts is always what you see in the respective area.
Value and completeness on the cards
Each card in the analysis hub shows two different things: completeness (is something filled in?) and the value from the analysis (how good is it?).
Good to know: A low value is not a verdict on the quality of your work but a statement about how much room for improvement the analysis found. The trend over time says more than the absolute value.
Overall assessment
Condenses the individual findings of the project into an overall picture – it is assembled, not freely written.
Good to know: Because it is assembled, it changes only when the underlying findings change. "Soundly structured" means: nothing conspicuous was found – not: reviewed on the merits and found good.
Rate presentation
A quick outside view of an uploaded presentation, without any project having to exist.
Good to know: It deliberately names only what is weak, not how to do it better. It reads text, not design. With extensive files only the beginning is assessed.
Argumentation – Building a persuasive structure
This is where the common thread takes shape: the introduction, the core statement, the structure in three chapters and the arguments that carry it. The functions build on one another – changing something further up changes everything below it.
Intro (Situation – Challenge – Decisive Question)
Produces the three introductory elements: the shared starting point, the challenge arising from it, and the question your presentation answers.
This improves the result: A clearly described recipient role shapes perspective and address the most. Whether you speak internally or externally changes the direct form of address in these three elements.
Good to know: Figures deliberately disappear from the intro – it should establish the shared starting point, not already prove it. The decisive question is the bottleneck: it determines what the whole argumentation is heading towards.
Key message
Formulates the one sentence your audience should take away – as an answer to the decisive question, not as a summary of the topic.
This improves the result: The core message hangs directly on the decisive question. To sharpen it, sharpen the question first.
Good to know: There are two routes to a core message: the analysis run and the suggestion button in the story area. They may word it differently – that is not an error, but two attempts at the same task.
Storyline structure
Selects from a curated library of three-part argumentation patterns the one that fits your undertaking, and derives the three chapter titles from it. Via "Suggestions" you get several patterns to choose from.
This improves the result: Impact goal, difficulty and a sharpened core message lead to a better-fitting structure. In any case you keep the last word: every pattern offered is freely selectable.
Good to know: The chapter titles are not messages – they are short labels. The messages are created afterwards.
Connected or unconnected structure
Before the pattern is chosen, the application decides whether your chapters should build on one another logically (connected) or stand side by side as three equally ranked topics (unconnected).
Good to know: Only three entries from the briefing are decisive: the training checkbox, the impact goal and the difficulty. A text that happens to sound like training does not trigger it – the checkbox does.
Chapter messages
Condenses the arguments of a chapter into the one statement the chapter carries – aimed at the core message.
Good to know: This stage cannot add anything: it summarises what the arguments contain. A weak chapter message is usually a sign of weak arguments, not of weak wording.
Argument topics
Derives which topics your argumentation must cover at all – the basis from which the individual arguments are then created.
This improves the result: The project description is the main source; precision pays off most here. Roles in the briefing help determine which supporting and which contentious topics are found in the first place.
Good to know: A topic not found here cannot show up as an argument later. If you want more arguments, you need more topics – and their number follows from the time frame.
Argument messages
For each topic, formulates the statement ("So what?") instead of only the subject ("What?") – with a rationale that draws on your sources.
This improves the result: Well-maintained materials are the factual basis of the rationales. Meaningful chapter titles improve how the arguments are assigned.
Good to know: The number of arguments is tied to the number of topics, and that in turn to your time frame. A chapter does not get more arguments simply because one would like them.
Optimise messages
Sharpens an existing message linguistically without changing what it says.
Good to know: "Without changing the content" is an instruction to the language model, not a check. Read the result against the original. The optimisation sees the one message, not its neighbours – repetitions between chapters escape it.
Rhetorical questions
Turns sensitive messages into questions so that an uncomfortable statement does not stand bluntly in the room. Three levels from neutral to provocative.
This improves the result: The difficulty determines which level is recommended. Whether you speak internally or externally, and which form of address applies, carries into the question itself.
Good to know: The selection is deliberately sparing: if everything becomes a question, nothing has an effect any more. The recommendation is a suggestion – you decide message by message.
When messages become questions
Two separate rules determine where the application offers a rhetorical question at all: one for the chapter level, one for the individual arguments.
Good to know: Because there are two rules, the settings act differently: the difficulty influences one, the impact goal the other. The training checkbox overrides both.
Background information
For each argument, the evidence and reasoning behind it – what needs proving, which objection it addresses, what would be lost if the argument were missing.
Good to know: What you enter here counts as established ground for all later steps: slide texts, speaker notes and Q&A answers draw on it. Entering something unverified here spreads it across the whole presentation.
Tools on the argument
Merge, split or sort out arguments – the application suggests which belong together and which contribute little.
Good to know: These tools only suggest; nothing changes until you agree. Merged arguments cannot be separated again at the push of a button.
Logic analysis
Checks whether the arguments carry their chapter message and the chapters carry the core message – and whether the whole fits the audience.
Good to know: The point values are assessments by a language model, not measurements. Two runs can turn out slightly differently.
MECE Analysis
Examines the arguments from two sides: do they overlap, and do they together cover what is needed?
Good to know: "Overlap" does not mean "error" – some repetition is intended. The analysis sees text, not intent.
Consistency check
Checks whether the central fields fit together – above all whether the core message and the decisive question answer each other.
Good to know: Similarity is not correctness: two fields can match well and both be wrong. An empty field is not a bad field.
Tools for the Arguments (One-Pager)
The one-pager features two tools side by side, each serving a different purpose. The checking tools identify weaknesses in the content – weak, duplicated or redundant arguments – and suggest amendments. The scaler, on the other hand, works towards achieving a desired number of slides.
This improves the result: The checking tools are useful if the substance of your argument is already in place and you want to refine it. The scaler is useful if a specific length requirement has been set.
Good to know: Each tool displays its suggestions individually for you to accept. If you do not accept a suggestion, nothing changes – and you can reset the selections at any time.
Argument Map
Shows which aspects your audience expects and which of them your arguments already cover – that is, where a gap remains.
Good to know: "Covered" means that an argument matches an aspect linguistically – not that it deals with it convincingly. The map is only as current as the case analysis it rests on.
Explore variants
You change individual inputs of the briefing – such as impact goal, difficulty, recipient roles, tone or duration – and have the argumentation recalculated with them. The result appears next to your current state: key message, chapters and arguments side by side, plus both argument maps and a short summary of what your change shifted.
This improves the result: Change only one or two inputs per variant – otherwise you cannot tell which change made the difference. Project description, impact goal and recipient roles shift the argumentation most; tone and addressing style mainly change wording, which the comparison of the structure hardly shows. If you change the recipient role, adjust the recipient description as well. “Recalculate baseline” shows what differs without any change; under “Calculated variants” several runs are shown side by side, and that is where you name and delete variants.
Good to know: Your project stays unchanged until you apply something: individual fields, everything or – after a confirmation prompt – the changed inputs to the briefing. Two runs with the same inputs never turn out exactly alike, so not every difference comes from your change – “Recalculate baseline” makes this share visible. The button is visible only to the owner of the project and only if the account includes the feature.
The guide checks along
When you add or change an argument in guided mode, the guide reports whether that holds up methodically – for example because a chapter becomes overloaded or two arguments overlap.
Good to know: "Contradicts" is an objection, not a block. You carry out every change anyway; the application only tells you what it thinks of it.
Presentation – Create and export media
The argumentation becomes slides, notes, visualisations and accompanying documents. Almost everything here depends on the quality of the argument messages – improving them improves everything downstream.
Slide text
Produces the text of a slide from the argument assigned to it – aligned to the message, not to the topic.
This improves the result: Argument message and background information are the strongest levers. Assigned charts and tables bring real values that may be used. Layout, audience and tone of voice also have an effect.
Good to know: "Too few lines" is usually intended: a slide should carry, not contain everything. And a figure that appears neither in the background information nor in an assigned content block is deliberately not invented.
Condense slide texts
Shortens the text of a slide – and with strong condensation it is not merely shortened but regrouped across the available text containers. Also possible for all slides at once.
This improves the result: How many text containers the slide has determines how many topics can be distributed.
Good to know: Strong condensation cannot be undone; the original text is gone afterwards. That individual containers then stay empty is intended. Images and charts remain untouched.
Speaker notes
Produces what you say about the slide: the lead-in, the explanation, the transition to the next slide – not the slide text in prose, but what it does not show.
This improves the result: Settle the slide text first, then generate the notes. The form of address takes effect here directly, as do tone of voice and language.
Good to know: Reordering arguments makes all the transitions around them wrong – after a rearrangement, have the notes generated again.
Visualisation recommendation
Suggests a suitable form of presentation for a slide – a chart, a diagram, a comparison.
This improves the result: Write the slide text first, then fetch the recommendation: with text the signal is considerably stronger. Without assigned data, data-driven charts drop out.
Good to know: The recommendation is a suggestion, not an automatism. And preview and export do not always display visualisations identically.
Assign content blocks to slides
Distributes the content blocks obtained from your sources – texts, images, charts, tables – to the arguments they fit in terms of content.
Good to know: What is compared are labels, not contents: a precisely named argument gets better matches. That an image does not appear everywhere is intended – images are distributed more sparingly than text. An argument without a content block is often the correct result.
Optimise slide design
Rearranges the elements of a slide or a SlideDoc so that no gaps remain and the arrangement matches the number and type of contents.
This improves the result: You first see a preview; nothing changes until you agree.
Good to know: Merged containers cannot be separated again at the push of a button. With a corporate template that has strict design rules, some freedoms deliberately stay locked.
Q&A preparation
Collects the questions you must expect, prioritises them and formulates answers from your own documents.
This improves the result: Well-maintained background information and evidence attached to the arguments are the factual basis of the answers – and spare you the follow-up questions about where a figure comes from. The challenges from the briefing are a second search direction.
Good to know: "Not apparent from the available data" is an intended result, not a malfunction – better an honest gap than an invented answer. The questions age: if you change the argumentation, generate them again.
Scaler – shorten and extend
Brings the presentation to a target number of slides or a target duration by summarising or supplementing in stages – and flags which summarisation is risky.
This improves the result: The better maintained the slide texts and background information, the better the pairs proposed for merging.
Good to know: The scaler asks before it acts. "Risky" is meant seriously – substance is lost there when summarising. And shortening cannot be undone by extending afterwards.
Template Mapper
Link the content of your presentation to the placeholders in your PowerPoint template. You define once and for all which layout is to be used for the title, introduction, agenda and contents slides, and which placeholder is to contain which content. Thereafter, all slides will be created in accordance with your design.
This improves the result: ‘Auto-Suggestion’ generates an initial suggestion; the preview next to it immediately shows what a slide would look like with it. If your template is missing a field for which CAai provides two options, combine them into a composite field.
Good to know: This assignment applies to your account. If someone else uses the same template, their assignment remains unaffected – and vice versa.
Saved language versions
Translated versions of a project are retained. If you later switch back to a language that has already been generated, you can use the saved version instead of having it translated again.
Good to know: If you have made any changes to the content since the version was produced, it is now out of date in those sections. CAai will then retranslate those sections rather than leaving the old version as it is.
The four text formats
Alongside the presentation, standalone documents are created from the same content – for instance a one-page summary or a fully written-out text.
This improves the result: The chapter titles become section headings – meaningful titles pay off especially here. The tone of voice can be chosen separately for each format.
Good to know: The chapter messages appear verbatim in the text. An unfinished message stands out more here than on a slide.
Applies everywhere
These functions are not tied to any phase – you encounter them in many places in the application.
Suggestions (compass button)
The compass next to an input field generates a suggestion for exactly that field, taking into account everything you have already filled in.
This improves the result: The more context is available, the higher the confidence level shown for the suggestion. A suggestion right at the start is by its nature an offer, not a recommendation.
Reasons behind suggestions
Many recommendations carry a reason why they make sense. In the analysis hub it sits below the suggestion and can be expanded; the "Show reasons" switch turns it on and off.
Good to know: It explains the recommendation plausibly – but it does not record which rule actually applied in the individual case.
Change language, form of address and tone of voice
You can change a project's content language, switch between formal and informal address as well as singular and plural, and change the tone of voice. Existing texts are carried over and adapted.
This improves the result: A project's content language is separate from the language of the user interface. All generators follow the content language – not the language you happen to be operating the application in.
Good to know: The change runs across all fields and takes a moment. Prepared questions and answers are translated separately.
Notice "Settings changed"
If you have changed fundamentals – such as the impact goal, difficulty or audience – the application points out that content already generated still rests on the previous state.
Good to know: This is a notice, not an error message and not a demand. You decide whether and what you have regenerated.
Share the presentation on feedback
You create a link that allows anyone to view and comment on your presentation – without needing an account or access to your project.
Good to know: Feedback submitted via a sharing link is treated as suggestions. It is up to you to decide which of these will be incorporated into your project.
The four permission levels of a share link
“View only” and “Comment” show your presentation in a reading view. “Edit” additionally allows changes to texts – either taking effect immediately or only once you accept them. All three require no account and work on your project.
Good to know: Set a password if the link might be passed on. You can revoke any link at any time.
Demo mode – the fourth level
This level differs from the other three: the invitee creates their own account, valid for 30 days, and receives a complete copy of your presentation. In it they operate the full app and may try out selected AI functions a limited number of times – analysing once, generating one overall proposal, and so on. It is meant for showing CAai to someone.
Good to know: They cannot change anything in your project – they work in their copy, and after 30 days both account and copy are deleted. They cannot download anything: export is blocked. Do note, however, that they get to see your content in full.
Working together on a project
You can invite a colleague to collaborate on a project. You will then both be working on the same content.
Good to know: Changes to the structure of the argument remain your responsibility: staff suggest them, but you are the one who implements them. You may edit content such as slide text and speech notes directly.
Team – Making projects visible
If a team consists of several people, they can see each other’s projects and can view or duplicate them to build on them.
Good to know: This visibility setting applies to all your projects, not just individual ones. If there is a project you do not want anyone to see, mark it as private.
Exporting and importing project packages
A project, including its sources, content modules and media, can be exported as a package and imported elsewhere – even into a different account.
This improves the result: The usual way of handing over a project to another person or to a different environment without losing any content.
Good to know: When you import the data, a new project is created in your account. Folder assignments from the source are deliberately not carried over – you organise the project yourself.
Filing: Folders and the Recycle Bin
Projects can be organised into folders. Deleted projects are first moved to the Recycle Bin and can be restored from there.
Good to know: The Recycle Bin is automatically emptied after a certain period of time. A duplicate is deliberately not assigned to any folder – even if the original is in one.
Roles and industries: free text and suggestions by the app
Roles of the audience and the presenter as well as industries can be typed freely or chosen from the catalogue – or both together. Whatever you enter reaches every stage: case analysis, overall proposal, storyline, arguments, slide texts and Q&A.
This improves the result: A precise free text (“hospital management and chief physicians”) carries more than an approximate catalogue role. Add the catalogue only if it really fits.
Good to know: If a field stays empty, the app suggests a role or industry after the first run. A suggestion is marked as such (light bulb, “Suggestion”) and is never written over your input – check, adopt or replace it.
Creating accounts in bulk and defaults for new accounts
Administrators create many accounts at once from an XLSX or CSV file. An organisation (tenant) can define defaults for new accounts: interface and content language, form of address, tonality and default template.
This improves the result: The file template contains all columns and permitted values. Values in the row override the default of the run; whatever is not specified comes from the tenant, otherwise from the app default.
Good to know: Defaults only affect accounts created afterwards. Existing accounts remain unchanged, and every person can change their own settings.
Page images of large PDF sources
For PDF sources, the app creates page images so that you can view pages and crop content blocks from them. For documents with more than one hundred pages, the images are created in blocks of one hundred pages – the first hundred first, further ones as you browse.
Good to know: If you browse to a page beyond the first block, the notice “Page images are being rendered” appears. This takes a moment; the text of the pages has long been read in regardless.
Quick tools and short argumentation
Small tools for text you enter directly – without a project, without a briefing. Useful for a quick formulation in between.
Good to know: The quick tools do not know your project: the text you enter is the only basis, and no fact-checking takes place.
Frequently Asked Questions (FAQ)
What’s the best way to start a new project?
Start with the briefing in Phase 1: define your target audience, objective and context. Then upload relevant sources. Based on this information, suggestions for all other areas can then be generated. Alternatively, use the Accelerator, which guides you through the entire process by asking key questions.
What is the difference between Accelerator and Guidance?
The Accelerator poses key structured questions and automatically fills in the fields – ideal for getting started quickly. The Guidance is an interactive dialogue that analyses your existing work and makes targeted suggestions for improvement – ideal for optimisation.
How do I improve my overall score?
Focus on the modules with the lowest scores. The analysis will show you prioritised recommendations. The most common areas for improvement are: a more comprehensive briefing, a stronger core message, and more and better sources.
Can I edit the suggestions?
Yes, all suggestions can be edited before they are adopted. You can also reject suggestions and write your own text. The suggestions are intended to support you, not to replace your expertise.
Will my data be used for AI training?
No. Your content is processed exclusively to generate your suggestions and is not stored afterward. The AI models are not trained using your data.
What do the levels of difficulty mean?
Difficulty describes how demanding it is to reach your impact goal with this audience – on five levels: “Simple”, “Rather simple”, “Medium”, “Somewhat challenging” and “Demanding”. It results from the sender’s credibility, the audience’s familiarity with the topic and its attitude towards the key message. Together with the impact goal, it determines the structure: with the impact goal “Inform” and “Simple” or “Rather simple”, the structure stays unconnected (organised by topic); in all other cases it becomes connected and deductive. If the training checkbox is ticked, the structure always stays unconnected. The higher the level, the more objections, evidence and safeguards the argumentation includes. The app also offers sensitive messages as rhetorical questions: with the impact goal “Trigger a decision/action” and “Somewhat challenging” or “Demanding”, and with the impact goal “Convince” and “Demanding”. If the audience is top management (e.g. board of directors or executive board), the app offers them only with “Trigger a decision/action” and “Demanding”.
What is MECE and why is it important?
MECE stands for ‘Mutually Exclusive, Collectively Exhaustive’ – without overlap and comprehensive. It means that your arguments cover all relevant aspects without overlapping. This is a hallmark of professional presentations based on the storyboard approach.
How does the Export work?
In Phase 3, you export your presentation as a PowerPoint, Word or PDF file – complete with a professional template and your Storyline structure, arguments and media. In addition, you can export one-pagers, speaking notes and Q&As as separate documents, and optionally attach background information and comments to the slide deck as supplementary slides.