There are cases when I delegate a high business value task to Jules, and that is usually when I wouldn’t be able to do it otherwise. The result might not be great, but you would not do it anyway so anything good that comes out of it is a benefit. For me it is a natural thing because business value will always trump technical certainty if those parameters have the same level. The only reason for that is because I care a lot about this result (high business value) so I want to oversee it personally and want to make sure it is ready as fast as possible.
It can free up customer service representatives to work with customers on higher-level problems. Organizations can use AI workflows to better manage the customer process, from onboarding new customers to sending them information about their purchase. IBM watsonx Orchestrate helps organizations create personalized AI agents to automate and accelerate their work—as well as providing an system to orchestrate complex workflows. Created by Open AI, ChatGPT is a chatbot that is widely credited with starting the gen AI revolution. It has several use cases, including inbound optimization, sales engagement and CRM improvements.
Here would be one case for using the Gemini CLI in YOLO mode, but we actually have a better tool for that called Jules. This makes your life much easier, because if you already know the steps, instead of doing them yourself you can simply instruct the AI to do it for you. The main reason to use the CLI for low technical certainty problems is that the feedback loop closes almost immediately. It takes a few tries for me to get the prompt right, in the same way it might take you a few tries to prototype something in a satisfying way coding it manually. Of course, the natural use case of the CLI is coding, but you know, people. Problems with lower technical certainty require more supervision to get things done the right way, so I prefer to use an interactive process with a CLI tool.
Components of AI workflow automation
In AI workflow automation, multi-agent architectures multiple specialized agents to collaborate, each operating concurrently under an orchestrating supervisor agent. These systems are capable of handling several interconnected processes with minimal intervention. In these structured sequences, AI systems perform, coordinate or enhance processes—either autonomously or in collaboration with human workers. While you can try to force Gemini CLI or Jules to do non-coding tasks, it feels like using a hammer to fix a screw to the wall. In summary — if my options are making zero progress or firing up Jules, I will fire up Jules to do some work for me.
Business value is how important this feature is to achieve the goals of the team. It involves gathering the relevant people together in a room for a discussion — including both engineers and stakeholders — and producing a logical sequence for the backlog to be executed. This is a method I’ve learned while working at ThoughtWorks in Agile transformation projects and then adapted to suit my own needs. That said, this scene is evolving every day and I believe it will only get better.
Solo developer
Given that, the natural choice would be to have a short feedback loop, so that for every change the AI proposes I can see the results immediately and guide them to the right https://www.internetling.com/speed-up-with-modern-programming-tools.html direction. This is a problem with low technical certainty, because I didn’t know which technology to use nor the exact places in code I had to change. For example, in this very blog I wrote last time how I implemented the “featured post” in the home page. If I am “invested” in a feature, I’ll prioritise working with it in a synchronous manner, but if not, I can delegate it to an async worker. High business value implies that the feature is critical for the business success, and low business value is likely to be a “nice to have” kind of feature.
- You define steps with typed input/output schemas using Zod, connect them with .then() for sequential chaining, use .parallel() for fan-out execution across multiple branches, and add conditional routing with .branch().
- AI powered workflows also merge duplicate records, append purchasing history, and surface upsell opportunities based on usage patterns.
- Developer-focused frameworks give you typed step definitions, branching and merging primitives, suspend/resume for human-in-the-loop patterns, and integrated tracing.
- Lead scoring workflows powered by LLMs can analyze behavioral signals, conversation history, and firmographic data together, something rule-based scoring can’t do.
- Sentiment analysis on incoming messages flags at-risk customers for proactive outreach before a complaint escalates.
For some customers, an AI-driven workflow that provides intuitive tools helps provide answers without needing to talk to a human, improving resolution speed and customer satisfaction. Organizations that created AI-driven, automated workflows are likely to be more efficient than those organizations that rely on more manual processes. Those employees can focus on high-value projects and tasks that drive extra revenue. Recently, for example, IBM tripled its number of open entry-level positions, planning to train early career workers in more intuitive, human-specific skills. AI frees up human workers to devote more time to customers or partners—as well as orienting more jobs toward service- or relationship-based positions. Business process automation (BPA) is a strategy that uses software to automate complex and repetitive business processes.
Low/Mid Technical Certainty OR High Business Value = Interactive Mode (Sync)
- When you wire a model call into a repeatable process with defined inputs, outputs, and orchestration logic, you’re building an AI workflow.
- Code generation has been the first breakout use case for developers but it is not the end of the story.
- 93% of developers use AI tools and nearly 30% of code is AI-authored source.
- A more complex version chains multiple model calls with branching, parallel execution, and human review checkpoints.
This product helps organizations identify leads and turn them into sales through AI-driven engagement workflows. There are several prominent tools and workflow automation platforms that use AI https://www.riverstonenetworks.com/why-ecommerce-development-is-growing-in-dubai.html to create advanced and automated workflows. AI workflows can easily route information and processes across the organization so executives and employees have real-time information wherever they need to access it.
You might have noticed that based on the categories above I singled out that high technical certainty with high business value is always synchronous. The important note is that because the feedback loop is long — you will not get the result immediately — https://stephanis.info/2019/11/17/if-you-think-you-get-then-this-might-change-your-mind-3/ it is better to reserve this tool for tasks that you have clarity of what to do step by step. Jules is an assynchronous agent that you can connect to GitHub to perform tasks in the background. Jules was launched by Google during I/O earlier this year, and quickly became my favourite tool ever (Gemini CLI comes as a close second).