Reusable AI: Raise the treasure — Part 3
AI use cases show great diversity. Not every business problem a priori has an unknown solution and requires research and scientific work. Many companies have similar problems, and these can be solved with reusable AI applications. AI can be like cooking.
In our previous article, we looked into the survey results of renowned analysts with respect to AI in business. Their findings can be viewed as top-level features of a future AI business platform:
- Reusable ingredients for AI solutions
- AI recipes for rapid response to recurring and similar problems
- Fast implementation through code-free orchestration
- “Predict as you train” paradigm for seamless operationalization
- Case management for a 360° view on AI solutions
- AI white boxing with comprehensible recipes to install trust
- Rapid solution deployment through dual use platforms
In this article, we introduce a next-generation AI business platform that implements these features right now: PredictiveWorks.
PredictiveWorks. is an AI catalyst and business-ready AI working horse with the mission to establish reusable AI. The focus is to boost small DataOps teams with code-free and reusable AI recipes for a wide variety of business use cases.
What fast.ai claims for deep learning, “making neural nets uncool again”, is our claim for the full spectrum of machine intelligence: We want to make machine intelligence as accessible as possible for business teams in companies who are not listed as Fortune 500s.
Before we continue to discuss PredictiveWorks. in more detail, we want to point out (again) that data processing is organized as a structured process, and AI is an important contributor to this process:
A standard data process comprises the phases: track, collect, aggregate, analyze, optimize and action.
AI is key to reach the action phase, but it is definitely not made to cover all phases of the entire data process. An important, but often neglected consequence: Corporate AI adoption must be part of a data strategy, and AI platforms complement platforms that track, collect and aggregate data.
Suppose a small or midsize manufacturer aims to optimize resource consumption or predict the next best maintenance interval:
Trying to directly connect an AI platform to a variety of sensors to compute insights from noisy raw data and ignoring that great IoT platforms are made for just this task, reveals a missing data strategy. With an expensive but avoidable effort to reinvent the wheel again and again. This argument holds for most of AI’s business use cases and shifts the focus from raw data to seamlessly connecting to appropriate data platforms that create the basis for AI applications.
PredictiveWorks. complements Google CDAP’s built-in data connectors for a wide variety of popular cloud services by 50+ purpose-built connectors from Cyber Defense and Internet-of-Things to E-Commerce and Marketing platforms.
Business-ready AI platforms must not restrict to the machine intelligence level. It is important that they provide an end-to-end support from business problem to associated tasks and down to AI workflows.
PredictiveWorks. connects dynamic case and AI workflow management with a 4-level AI taxonomy: AI projects receive a structured semantic guidance and every AI solution is transformed into a 4-layer business object — machine readable and human understandable.
It is an all-in-one configurable information package that describes the entire architecture of an AI solution, from business case down to machine readable instructions for a code generator to build executable AI workflows with a click. At any time, business users find answers why a certain data connector and operator has been selected, how they fit to a certain project task and what all of this has to do with the originating problem. Because of their configurable nature, these business objects represent and are named AI solution templates.
Next, we give a brief overview of PredictiveWorks.’ AI taxonomy.
The first taxonomy level is represented by a business case. Cases, adopted from dynamic case management, record the high-level business context:
The specific reason for the AI solution, why the specific approach is important for the project and more. Business cases prepare the basis to reuse projects for similar problems, accelerate the onboarding process of recently hired employees and more.
The second level is defined by business tasks to organize business cases into more detailed tasks. Business tasks can be categorized along proven phases of analytics processes: It is the CRoss- Industry- Standard- Process for Data Mining on the one hand and the Analytics Solutions Unified Methods for Data M ining & Predictive Analytics on the other hand.
We support five phases: Understand, Prepare, Build, Evaluate & Operate. Solving a business problem usually starts top down and with tasks to understand the problem and derive ideas how to solve it.
The third taxonomy level is defined by blueprints. A blueprint defines the logical plan of an AI application and determines which components are needed, how they interact and how they have to be configured to turn low value data into high value insights.
Blueprints are categorized by two semantic dimensions:
- Analytics. Descriptive, Diagnostic, Declarative, Predictive and Prescriptive
- Purpose. Discover, Ingest, Learn and Predict.
Instead of writing software to implement AI applications the traditional way, PredictiveWorks. abstracts from software engineering and shifts the focus to application design.
More than 200 pre-built & reusable Lego-like building blocks (plugins) exist, that can be configured and plugged together with a point-and-click data workflow designer to specify logical plans (blueprints). An application generator then turns every blueprint into an executable AI application. Plugins define the fourth level of PredictiveWorks. AI taxonomy. They can be categorized as Source or Sink of a data workflow, as Action, Alert, Analytics, Condition, Lookup & Transform.
The image illustrates how PredictiveWorks. exposes the DNA of a botnet detection solution in terms of business case, tasks, blueprints and pre-built plugins. Comprehensible and reusable business knowledge.
PredictiveWorks. responds to every business problem with a structured AI solution template organized along a 4-layer architecture. It is an important simplification for business:
Instead of managing hundreds of platforms and services, a single platform with a single technology, learning curve and user experience remains to be operated for many different use cases.
So, AI solution templates move into the spotlight as the next real business assets when it comes to transform every business process into a data-centric process. They can be registered and made available in a knowledge hub, sent by email or otherwise distributed. Easy access to best practice AI recipes is an important step to facilitate AI adoption. AI templates, however, do more: They describe, explain and expose — from problem down to AI workflows — every stage of an AI solution.
So, business users understand what their solutions do with their data — step by step. It is the opposite of magic black boxes, installs trust and further increases acceptance in business to use AI for human decision making.
PredictiveWorks. provides a template (knowledge) hub for a wide variety of reusable business cases, covering Cyber Defense, Internet of Things, E-Commerce and Marketing. This hub integrates a template recommendation and search engine and ships with all ingredients to offer the widely accepted user experience of a marketplace.
Boost AI Projects
Accelerating AI projects and relying on a resilient business process are key to reach millions of smaller companies. The implementation of the magic formula “enable few to reach millions” defines PredictiveWorks.’ traction and contributes to a sustainable democratization of AI.
AI projects will be organized along the following process and trigger the future AI wave:
1. Understand Problem
Knowing the business problem, its symptoms, drivers, potential risks and also desired outcome defines the starting point for any solution strategy, and it is often this initial step that slows down AI projects from the very beginning.
PredictiveWorks. template hub is organized as a marketplace with smart search and recommendation support, to help users get inspired by potentially similar symptoms, drivers and more. So, tedious and lengthy meetings are avoided, and AI projects gain traction from the beginning as they start with the evaluation of a set of matching AI solution templates.
2. Select Template
AI recipes provide a 360° end-to-end view, from business problem down to appropriate workflows for training AI models and computing insights. This problem-to-solution white boxing enables solution providers and their customers to select the right template for the right problem fast.
3. Customize Template
Many SMEs face similar problems, but they are certainly not identical. You might need to add an extra business task and data sources as well as destinations may slightly differ. There also can be a need to add different data operators to cover other business rules.
PredictiveWorks. offers customization and orchestration of every AI solution template from modifying business case to business tasks and AI workflows without the need to write any line of code. This enables solution providers to thrill their customers project with a rapid approach right from the beginning.
4. Register Template
PredictiveWorks.’ mission is to make AI solutions reusable. On the one hand, this reflects the fact that not every business problem is unique and accompanied by an unknown solution that needs research and scientific work. On the other hand, it is the only meaningful strategy to reduce the size of expert teams and accelerate projects to minimize time to value for millions of smaller companies.
So, PredictiveWorks. registers every created or updated AI solution template in its template hub by default to ensure reusability and sharing across project and even business boundaries.
5. Generate Solution
AI solution templates represent best practice AI recipes and are important business assets to significantly accelerate AI projects. Templates are human understandable and machine readable. The latter means that they contain machine readable instructions to automatically turn AI workflows into executable software binaries — with PredictiveWorks. integrated code generator.
PredictiveWorks. was made to support solution providers with an AI working horse to respond to their customer needs fast. To finally reach this goal and have a customized AI solution available, a single step is left:
AI solution providers leverage the executable binaries of the high-level training workflow to create the respective AI model(s). PredictiveWorks. supports this step with a big data runtime environment to execute the training binaries. An integrated model management registers every model, including parameters, version and achieved accuracy.
6. Deploy Solution
PredictiveWorks. supports the “predict as you train” paradigm and guarantees fast operationalization and deployment. Business users will receive the same platform for predictions that was used for training purposes with trained and registered AI models and the executable binaries for the production workflow.
PredictiveWorks. is designed as a dual-use platform to seamlessly support both phases of AI solutions with the same platform and technology. Deploying AI solutions has never been easier and faster.
Making AI uncool
Todays’ corporate AI adoption is cumbersome. More than 85% of AI projects never make it into production and do not return their investments. Even worse, in Europe alone, more than 1.6 million SMEs wait for AI — a clear indication that there is no momentum in business.
PredictiveWorks. is an AI Catalyst and it is made to change the current AI process. The focus is on solution reusability, fast operationalization and deployment.
The key assets are AI solution templates that can be orchestrated and customized without writing a single line of code. Templates provided a structured end-to-end view on AI solutions from business problem down to technical AI workflows that are backed by Lego-like toolkit of 200+ prebuilt data connectors and operators.
It has never been easier for an AI solution provider to respond to customer needs with small DataOps teams and a minimal time-to-value. PredictiveWorks. makes building AI solution like cooking, with the right ingredients, the right recipes to generate the right AI dish for millions of waiting SMEs. Solution providers can perform significantly more AI projects in shorter periods of time.
This implements the magic formula “enable few to reach millions” and creates the missing AI momentum in business.
PredictiveWorks. is the basis for a reliable and structured process to turn business problems into reusable AI solution. That and the integration of main concepts of dynamic case management make AI projects uncool again. Agile project management methods can be used to plan and control AI projects in a given budget and time frame.
AI can be like cooking — ready to open an AI restaurant?
Originally published at https://www.linkedin.com.