In the age of digital transformation, businesses have faced a trade-off: Speed and scale often come at the cost of genuine human connection. But now, with the rise of AI agents, that equation is changing — it doesn’t have to be a choice.
Companies can move faster and get closer. Automate more and care more. The most successful companies deeply understand their customers’ needs, their journey, and their timing. AI agents enable them to provide the right help at the right moment, at scale.
This is the promise of the agentic enterprise — the next evolution of business, where trusted AI agents handle the busywork so your people can focus on what truly matters: Building trust, solving complex problems, and creating meaningful relationships.
The potential is so great, getting started can feel daunting. The most successful AI deployments don’t begin with code, but with a clear, strategic framework. Building an AI agent is more than just a technical task; it’s a strategic process that requires a comprehensive plan. It’s like writing a job description, architecting an IT system, onboarding a new employee, and building software all at the same time.
When I work with customers, I like to start with a five-step process to define key attributes that provide a clear blueprint for success — helping move from agentic ambition to results, with confidence.
The first and most crucial step is to decide what outcomes you want to drive and the agent’s role in achieving them — much like you would create a job description that outlines goals, swimlanes, tasks, and success metrics for a human employee. What business function will it serve — sales, marketing, or customer service? What are its ultimate goals? You’re giving your agent a mission statement and defining the business processes and logic it should use.
For example, a large healthcare company was grappling with how to handle a massive spike in customer inquiries during open enrollment, which overwhelms their human support teams every year. Training and scaling a human workforce to handle complex, nuanced support issues for only a short period of time isn’t practical, so they wanted to see if agentic AI could help.
First, they defined their agent’s role: to support their human workforce by handling common questions and guiding customers through the enrollment process, deciding when to escalate to a human to handle more complex cases. Having a clear vision of the agent’s role and purpose allowed the company to define what topics it should cover, and and then provide clear instructions on how to proceed.
Once the role is clear, you need to equip your agent with the right context. Just like human employees, an AI agent’s effectiveness hinges on the information it can access and learn from.
In the healthcare example, the agent needed to do more than just answer general FAQs. To provide personalized service, it required access to a wide range of data, from unstructured conversations between patients and healthcare providers, to structured customer health records and insurance policies — unified to give the agent a complete view of the patient. This ability to integrate, ingest, and process diverse data sources helped the agent generate answers specifically tailored to each patient. Without proper data, agents ultimately end up needing to hand off more cases to humans, greatly reducing the opportunity to maximize human workers’ potential.
With a defined role and access to the right data, you then want to empower your agent to act. This is where you determine its autonomous capabilities and the business processes it can execute.
For example, the healthcare agent can verify insurance coverage based on the patient’s specific plan and policy, share recommended appointment times and schedule them, and even make personalized recommendations for ongoing care based on a patient’s history.
The key is not to reinvent the wheel. First, focus on automating the time-consuming, repetitive tasks that already exist in your workflows, so your human employees can focus on higher-value work and spend more time with the patients themselves
An agent’s ability to handle complex and sensitive tasks — like managing health information — makes guardrails essential. By defining guardrails for your agents, you prevent unintended actions. What actions should they avoid, and which ones require additional oversight?
For the healthcare company, this meant establishing a clear protocol for when to hand off a task to a human. For example, the agent was programmed to identify potentially concerning symptom patterns and immediately escalate the case to a human physician for review. This gave the agent the autonomy to act but also the wisdom to know when to ask for help.
The final step is to determine where and how the agent will operate. Will it be a public-facing assistant on your website, a collaborator with human teams in a platform like Slack, or a back-end bot working with other agents and automating processes in the background?
For the healthcare company, this meant ensuring the agent operated exclusively within HIPAA-compliant systems, protecting sensitive patient data and building trust with both patients and providers.
These five attributes help ensure your agents get off to the right start, with clear goals, guidelines, and autonomy to be successful. The journey doesn’t end with a launch; it’s just the beginning.
Once deployed, evaluation, management, and ongoing adjustment are key to ensuring their ongoing success — just like human employees thrive with practical guidance and performance goals. As the agent proves its value and accelerates in its role, it can be “promoted” by revisiting these five steps to layer in new capabilities throughout. In doing so, you can unlock further potential within your workforce and create a deeper connection between employees and customers.
Platforms like Agentforce, coupled with the broader Salesforce ecosystem, provide the necessary tools to build and manage agents, unify data, and take action directly in the flow of work. This strategic, platform-driven approach enables organizations to transform ambitious AI aspirations into tangible business results, thereby accelerating their journey toward becoming a fully agentic enterprise.
YOUTUBE.COM/THENEWSTACK
Tech moves fast, don’t miss an episode. Subscribe to our YouTube
channel to stream all our podcasts, interviews, demos, and more.
Major lenders are building artificial intelligence-powered “agents” – software that can do the same work as humans – in their business banking divisions, as the battle for AI supremacy in financial services intensifies despite workforce concerns about the risk to jobs.
Commonwealth Bank of Australia is building what it describes as “virtual relationship managers” in its business bank. The customer-facing technology is in a pilot stage as the bank discusses the timing of a market rollout with regulators.
Loading…
If hard skills are increasingly being automated, employers are shifting focus to what AI can’t replicate: creativity, empathy, critical thinking, and other essential soft skills.
For years, technical abilities were king, but the tide may be turning.
Indeed’s Hiring Lab took a look at job postings and analyzed which soft skills were listed. The top were communication, leadership, and organizational prowess. Forty-three percent of all job listings had at least one soft skill advertised.
Soft skills show up in job postings across industries, but maybe not where you’d expect:
In a world where machines can write code and analyze spreadsheets, the need for human insight, emotional intelligence, and creativity has never been more critical.
Employers don’t just want workers who can do the job; they want people who can collaborate, innovate, and lead.
Sign up for BI’s Tech Memo newsletter here. Reach out to me via email at abarr@businessinsider.com.
Archie Mayani is the chief product officer at GHX, a global supply chain company that uses data and cloud-based technologies to connect healthcare providers like hospital systems and their suppliers.
For more than 20 years, Mayani has worked on clinical and supply-chain health technologies at companies like Change Healthcare and United Health Group.
At GHX, Mayani works to ensure that the company develops technology that can help hospitals procure patient supplies — like implants and IV fluids — as seamlessly as possible. By using AI-powered technologies that can anticipate supply chain disruptions, prioritize them in order of most critical, and identify substitutions, hospitals can be better equipped to provide effective patient care.
Business Insider interviewed Mayani about what sets healthcare apart from other industries when it comes to AI implementation.
This interview has been edited for length and clarity.
Rachel Somerstein: How is healthcare unique as an industry, particularly when we think about the integration of AI?
Archie Mayani: I’m based in Silicon Valley, where everybody wants to fail fast and move forward. But healthcare is very different from other sectors using AI.
When you are building a dating app and your AI hallucinates, it’s kind of funny and makes a great first-date story. When you have a patient on the operating table and you don’t have the right supplies delivered at the right time, it’s scary.
Can you talk about the goals of AI implementation in healthcare supply chain management?
Healthcare is about patient safety and how you use technologies responsibly, always putting the patient in the center. When we think about supply chain management, it’s almost like an invisible operating system in this shared ecosystem of patient care and delivery.
GHX’s mission with AI implementation revolves around delivering the right supplies at the right time to improve the quality of care and make it more affordable.
How did you arrive where you are now?
We have been leveraging AI and machine learning for the last 15 years. A lot of our work during the pandemic involved making supply disruptions more visible, with the goal of making supply chains more resilient and proactive.
One of the most important cases we thought about, coming fresh off the pandemic, was, “Can we look at backorder anticipation?”
It doesn’t matter what the cause is — it can be geopolitical conflict or meteorological tragedies. It could be that a trailer was dislodged and now we’ve lost the supplies on the freeway. But if we can anticipate back orders, we can anticipate disruption.
If the system is intelligent enough, it could recommend nearby substitutes within your distributed area. We started there, on a path of, “We’re going to build this machine-learning model that’s going to be intelligent, anticipate these disruptions, and make substitution recommendations.”
Where is AI in supply chain management working best right now?
We have an agile development approach at GHX, where our customers give us live feedback. We had an “aha” moment from our customers: They said, “This is absolutely what we’ve asked for for the past 20 years. You are starting to predict all of these disruptions, but the disruption of a Band-Aid is not the same as a disruption of IV fluid.”
They asked, “Can you make this technology even more intelligent for what I need, depending on where I think my most critical risks are and what kind of care delivery is most important to my organization?”
So we came up with the idea of clinical sensitivity and a confidence score, essentially to validate whether disruptions are clinically relevant to specific customers.
That was one of the things that changed the trajectory of our AI implementation road map: Just because we can deliver insights doesn’t make them useful; they have to be predictive and personalized.
What does the future of AI in healthcare supply chain management look like?
Since healthcare is different and unique from other industries, our approach is to automate workflows as much as possible using agents while keeping a human in the loop. Once the customer feels confident, we can start fully abstracting those workflows so that AI agents are handling them entirely.
The other place gaining traction is copilot environments. For example, we have a product called the perfect order dashboard, which marries data insights. A customer may say, “Show me the view of my world, of where the supplies are, of where I’m doing an exceptional job with my suppliers getting those supplies on time, making sure that the orders and invoices are paid on time, and show me all of the discrepancies.” Still, that’s not enough.
The copilot allows you to tell a story with that data, very similar to a ChatGPT-like experience: “Show me the top three defaulting suppliers not delivering supplies on time.”
Once you have those supplier lists generated, you can say, “Send an email to XYZ supplier, making sure we have a quarterly business review scheduled, and please attach the perfect order dashboard view showing the last quarter’s trend.”
It might seem small, but it’s a huge value-add. It used to take maybe three or four hours to understand the data, extract insights, and drive follow-up actions and decisions. Now, it takes minutes.
What advice do you have for others in your position or who hope to be?
The hardest or most useful thing you can do is to say no.
In healthcare, everything is urgent — and it truly is. But not everything matters equally. So, the ability to say no to the right things and ensure that you’re focusing on the highest value-added items for your customers is critical when you’re in healthcare.
Big Tech, or even a smaller tech startup, can innovate as research labs and fail. We don’t have that option. So understanding what matters now, what will matter in 10 years, and finding the right balance to focus on the right innovations, becomes critical.
It’s about having the right data, the right governance and mechanisms, and always thinking about performance, security, and privacy. It’s also about making responsible choices on where to invest your energy, so that you’re ultimately not working on the sexiest, coolest, or hardest things.
It comes back to the patient: making care affordable and of the highest quality possible.
The Guardian view on Trump and the Fed: independence is no substitute for accountability | Editorial
Building Trust in Military AI Starts with Opening the Black Box – War on the Rocks
SDAIA Supports Saudi Arabia’s Leadership in Shaping Global AI Ethics, Policy, and Research – وكالة الأنباء السعودية
Journey to 1000 models: Scaling Instagram’s recommendation system
Mumbai-based Perplexity Alternative Has 60k+ Users Without Funding
Happy 4th of July! 🎆 Made with Veo 3 in Gemini
Macron says UK and France have duty to tackle illegal migration ‘with humanity, solidarity and firmness’ – UK politics live | Politics
VEX Robotics launches AI-powered classroom robotics system
Kayak and Expedia race to build AI travel agents that turn social posts into itineraries
OpenAI 🤝 @teamganassi
