Custom AI chatbot implementation framework
Supplier landscape
‣
Other information
‣
Custom AI Chatbots
Custom AI Chatbots
Introduction, Use cases, Getting started
Christian Ullrich
Outline
- Introduction
- Implementation
- Getting started
- Use cases for custom AI chatbots
- See it in action
- Closing remarks
1. Introduction
What is artificial intelligence?
- AI is about making computers behave intelligently.
- Some keywords: Machine learning, neural networks, natural language processing
- AI performs tasks that need human intelligence: More than logical rules.
What is generative artificial intelligence?
- Generative AI creates new content: Writing, images, and melodies.
- Realistic, coherent outputs.
- ChatGPT is an example of generative AI: Generating relevant sentences based on input.
What is ChatGPT?
- ChatGPT is an AI software for human-like text based on input.
- Useful for brainstorming, document outlining, and content drafting.
- Not so useful for calculations, facts, and research.
How can we use ChatGPT?
- ChatGPT can draft emails, write code, write documents, and translate languages.
- Many tasks involving text generation.
- Challenge: Find your individual use cases.
What are information repositories?
- Information repositories store and maintain (written) information.
- Websites, internal wikis, knowledge base, research databases
- A broad term for many existing things.
Why do we need more public information repositories?
- Accessibility, transparency, accountability
- Making information available to anyone: Democratizing information.
- Political strategy: Data legislation, Freedom of information acts
How can we create information with the help of ChatGPT?
- ChatGPT reduces the amount of work to create high-quality written information.
- ChatGPT can rework existing low-quality content.
- ChatGPT can simplify complex and complicated content.
What are classical chatbots?
- Classical chatbots simulate conversations with humans based on predefined rules.
- They struggle to handle unexpected user inputs.
- Classical use case: Customer service to handle simple queries
What are custom AI chatbots?
- Internal information + ChatGPT capability.
- Specific sources (+ ChatGPT, if desired).
- New software category after ChatGPT was launched.
How can we use custom AI chatbots?
- Customer service, marketing, sales. And much more.
- Answer questions. Research trusted sources. Act as a 24/7 advisor.
- Like a classical chatbot but better, faster, (cheaper).
2. Implementation
How can we use ChatGPT for our organization?
- Research
- Brainstorming
- Creation
- Rework
- Preparation
How can we share more information?
- Change your mindset.
- Create more.
- Share more.
How can we use a custom AI chatbot?
- Search on public websites.
- Search through semi-public content. (e.g., magazines)
- Internal search. (Data privacy challenge)
- Build a research alliance.
How can we differentiate with a custom AI chatbot?
- Faster, better, cheaper information dissemination.
- New business models.
- Public relations. (”We use AI”)
- Learning.
3. Getting started
Try out ChatGPT!
- Experiment with ChatGPT.
- Try different prompts to understand its capabilities.
- Learn “prompt engineering”.
Share more information publicly!
- Identify information that can be shared publicly.
- Align your internal stakeholders. (Aaargh!)
- Share it.
Try a custom AI chatbot software tool!
- Set up a project.
- Align your stakeholders. (Aaargh!)
- Set up the technology in a day. (Literally!)
- Follow a methodology.
Work together!
- Try ChatGPT indivudally.
- Set up a “community of practice”.
- Share your experiences within the organization.
Get help!
- Don’t outsource.
- Build your methodology with the help of ChatGPT.
- Get a trusted advisor.
4. Use cases for custom AI chatbots
Generic use cases
Marketing & Sales
- Website search
- Customer education
- Information consolidation
Customer service
- Information already exists.
- Less-prepared internal content.
- 24/7 customer agent. (But actually working.)
Internal operations
- Brainstorming and writing
- Search & Advice
- Process execution based on natural language
Research available to the public
- Search at your organizational website.
- Research on websites of various organizations.
- Separate business models.
Information-rich service providers
- All professional services
- Digital business development
- Separate business models
Armed Forces
Training
- Share training material more openly (especially if created with ChatGPT).
- Offer trustworthy(!) information repositories.
- Make training material adaptable (but still correct and trustworthy).
Standard Operating Procedures
- Write additional practices and procedures with ChatGPT.
- Share them publicly.
- Create an AI showcase project.
Recruitment
- Offer a 24/7 recruitment office. (Working business case!)
- Offer educational information for high school students. (e.g., computer science courses) (role model: America's Army)
- Offer 24/7 online teaching on relevant topics.
Public relations
- Collect information repositories (and there are many).
- Make information “accessible” (and even more accessible).
- Last but not least: Showcase AI adoption to stakeholders (e.g., journalists).
Defense research
- Collect information repositories across organizational boundaries.
- Align stakeholders (Aargh!) across organizational boundaries.
- Showcase AI adoption in research.
Chambers of Commerce
a. Website search
- Website search is nearly always awful if it’s not based on Google.
- Custom AI chatbot as an addition, not a replacement.
- Interactive guidance instead of finding the right page.
b. Training & Education
- Educational content hub
- 24/7 teacher based on trusted sources.
- New business models together with other organizations.
c. Economic research
- Explanation instead of search.
- Complex content for new target groups.
- Research for practical guidance.
d. Legislative advocacy (aka Lobbying)
- Explanation instead of search.
- Explanation for more target groups.
- Joint lobbying with other organizations.
e. Business directories
- Business listings 2.0
- Find the right business without maintaining a directory.
- Make trustworthy educational information accessible.
Education
AI learning
- Courses about AI software tools and their use.
- Practical adoption of software tools.
- Learning the value of information sharing.
Teaching
- Trustworthy learning content.
- 24/7 support.
- AI showcase implementation.
Student content sharing
- A platform for sharing information created by students (with or without the help of ChatGPT).
- Re-using already created content by other classes.
- Teaching the sharing economy.
Inter-organizational information exchange
- Making information accessible without publishing them in one information repository.
- Building networks of trustworthy sources. (even competing views)
- Limited organizational alignment, limited “politics”.
New business models
- New business models based on open-source information.
- Offering information with self-service customization.
- Making already free information accessible.
IT management
a. Digital strategy
- Make the strategy “useful” through AI guidance.
- Discuss strategies openly.
- Showcase your AI implementation.
b. Standard operating procedures
- Make SOPs practical.
- Share SOPs openly.
- Develop a SOP platform.
c. User help & training
- Make guides practical.
- Offer 24/7 support.
- Make training material applicable.
d. Service catalog
- Publish public service catalog as AI showcase.
- Offer intelligent search.
- Brainstorm procurement business models.
e. Sourcing & Supplier management
- Improve contract (lifecycle) management.
- Share contract templates with the public.
- Help with contract writing based on trustworthy sources.
Health
Citizen information
- Find answers from a trustworthy source.
- Combine several trustworthy sources into a knowledge hub.
- 24/7 virtual paramedic.
Professional research
- Q&A for professionals.
- Education for professionals.
- Scientific research.
Policy & Regulation
- Citizens
- Citizen advisors
- User help desk
New business models
- Information dissemination
- Shift from medic to paramedic.
- Disruptive business models in emerging countries.
Inter-government cooperation
- Platforms for publicly funded information.
- Integrated translation.
- Public health policies & practices exchange without document exchange or translations.
Management frameworks
Practicability
- Applicable management information.
- Open frameworks instead of bullshit bingo books.
- Concise answers instead of 230 pages.
New frameworks
- Lean testing of new management approaches.
- No book, no training, no implementation.
- Build, share, wait.
Market demand
- Nothing new but constant renewal of existing concepts.
- New focus: Simplification & Increase of speed
- New target users: Students, Frontline workers, Specific professions
Human experiences
- Collection of actual experiences.
- Reports as basis for answers.
- No need to read long interviews.
New approaches
- New approaches to “framework writing”.
- Significantly faster development with ChatGPT.
- New business models. (Content can/will/must be free.)
Political communication
Self-service information
- Stakeholders pull information themselves.
- Stakeholders know how to use information.
- Stakeholders need less manual customization.
New target groups
- Reaching new target groups with less effort.
- Less information customization.
- Quicker target group trials.
Inter-organizational cooperation
- Cooperation without a central information repository.
- Simple onboarding and offboarding of information repositories.
- Chatbot links sources: Website traffic generator.
Internationalization
- International cooperation with less effort (e.g., translations).
- Re-use of information resources (and even repositories).
- Collaboration on information development (with future customization).
New business models
- Private/Public funding sources allow new business model testing.
- Open work environment between organizations facilitates ideation.
- Evident need for efficiency gains in policy development.
Political education
Perfect use case
- Information shall be distributed openly and freely.
- Nearly no commercial or other restrictions.
- Inter-organizational cooperation is already happening.
Organizational cooperations
- Making information accessible without publishing them on one platform.
- Building networks of trustworthy sources. (even competing views)
- Limited organizational alignment, limited “politics”.
Self-service information
- Less teaching, more self-service learning.
- Less central customization necessary.
- Scalability of the political education model.
Simpler customization
- “Everyone can be a teacher.”
- AI-assisted self-service customization.
- Platform instead of pipe business model possible.
Improved accessibility
- Content is actually used.
- Possibly less but more intense use.
- Content use by more target groups.
Procurement
Process design
- Make your purchasing guideline accessible to suppliers.
- Publish process descriptions and SOPs.
- Develop industry standards.
Contract (lifecycle) management
- Make contracts accessible.
- Offer a 24/7 contract management office.
- Answer questions by using an AI chatbot but not making the chatbot accessible.
Statement of work
- Share your statements of work with other organizations.
- Offer a trustworthy(!) statement of work assistant.
- Integrate a chatbot for suppliers into your procurement process.
Supplier directory
- Crawl suppliers’ websites.
- Offer a supplier research chatbot.
- Develop a business model.
Advanced process support
- Integrate your internal apps with classical and custom AI chatbots.
- Execute processes with natural language.
- Improve processes based on prompt analysis.
5. See it in action
ChatGPT
- Try it yourself.
- Conduct a workshop.
- Share your experiences.
Information repositories
- Websites
- Scattered research
- Internal information to be published
Custom AI chatbot
- customgpt.com
- intrenion.com
- TBD…
Workshops
- ChatGPT
- Information management
- Custom AI chatbots
6. Closing remarks
Start simple!
- Try ChatGPT.
- Share your experience.
- Share your results.
Think from the goal!
- Where would a custom AI chatbot be useful?
- What information repository do we need for that?
- How can I start with ChatGPT?
Why generative AI?
- It works.
- It will change everything.
- It works for you. (After max. 30 min of brainstorming.)
Why now?
- It’s useful now.
- It’s accessible now.
- It’s affordable now.
Why you?
- Someone has to do it.
- Lead, follow, or get out of the way. (General Patton)
- Grow yourself, grow your career, grow your organization, grow humanity.
Why together?
- Think collaboratively. (It’s hard, I know.)
- Join forces.
- Inspire others.
Last remarks
- My humble opinion.
- Discussion is just starting.
- Other opinions are welcome.
Questions & Discussion
Christian Ullrich
intrenion.com
linkedin.com/christianullrich