Leading pharmaceutical companies are adopting tools like Botcopy and Google Cloud Dialogflow to create internal-facing chatbots. The goal: save time, save costs, and, most important of all, save lives.
Genentech is a pioneer in biomedical research, and a member of the Roche Group, one of the world’s largest pharmaceutical companies.
At Botcopy, we’re excited to share highlights from our interview with Genentech, below. As I write this in April of 2020, we’re in a pandemic, so fast-tracking clinical trials is on everyone’s mind. Perhaps you’ve wondered: What’s being done to speed up clinical trials? How is AI being used? Who’s in charge of implementing it?
Enter Jim Lucas, Systems Specialist, Genentech
Jim Lucas is one of the trailblazers bringing AI to clinical trials, intending to speed things up. He provides an insider perspective on what goes on in the clinical-trial trenches:
Jim Lucas, Genentech
Before getting a drug approved by the FDA, hospitals and clinics need to capture and record data while running clinical trials. To do this efficiently, clinical trial teams rely on different systems. We support ten to fifteen systems at Genentech.
Ten to fifteen systems seems like a lot. Especially since, according to Jim, each clinical system has its own moving pieces: multiple vendors, knowledge bases, venues for recording or retrieving info, and so on.
The challenge: creating simplicity out of complexity
Jim explains Genentech’s motives for building a clinical systems chatbot:
We wanted to marry all the different pieces together. Our company is large, so each department functions almost as a separate entity. Getting everyone in sync and getting budgets approved can be a process.
Fortunately, Jim was able to marshal the resources to get the clinical trial bot funded and deployed.
We’ve been able to move through what might have taken months or years in a quick amount of time.
Kudos to Jim and the people at Genentech for being uncommonly nimble. But even with administrative challenges overcome, there’s the not-so-small task of building a chatbot that’s demonstrably helpful and that people will use.
The solution: the right tools meet the proper techniques.
Genentech is using Google Cloud Dialogflow on the back end, and Botcopy on the front end. These tools used together offer Google’s leading NLU/ML framework, combined with Botcopy’s powerful web-based UI. These technologies are newer, so a lot of companies still don’t yet have specific staffers designated to execute on these projects. Dialogflow is HIPAA compliant and hence, has seen tremendous traction in the healthcare industry over the last couple of years.
There’s interest in leveraging AI and chatbots throughout the company. Our team wears multiple hats for now. I support systems, but since I understood the technology, I’ve been working on it.
Jim took on the unofficial role of “head of Google Cloud Dialogflow” within his department, while his managers helped move the project along with administrative work, budget approvals, and QA.
To help supplement Jim’s skillset, Botcopy’s conversational AI team helped Jim build out the responses within Google Cloud Dialogflow.
(Botcopy is a SaaS product that connects Google Cloud Dialogflow to websites, but we have an agency background, so we help build out agents on a case-by-case basis.)
Midway through the process, Jim’s project shifted into high gear due to unexpected developments. He explains:
Our initial use case was straight FAQs for our systems. But suddenly, with the pandemic, we noticed we had all these departmental task forces receiving emails or tickets, and a backlog was forming. We created Google sites to host searchable spreadsheets, but it was getting complicated.
An opportunity to save oceans of time. And possibly lives, too.
Jim did a few calculations and arrived at a stunning conclusion: The person writing the FAQs, responding to emails, and updating spreadsheets, took 14 minutes per question, per user. On average, it took seven hours for one person to manage issues for just one department. Now multiply seven hours by all the departments, and let’s say you have at least 70 hours per day of people doing things a bot could do. Multiply that by 365 days, and you have over 25,000 hours per year that could potentially be automated.
But that’s just the beginning of the equation.
With a chatbot, Jim realized Genentech would be saving significantly more than 25,000 hours. For example, time zones impact the time it takes to answer.
The person handling the support is in Australia. We’re a global firm, so if clinical teams are not in that time zone, they might have to wait a day or two for an answer. So on top of the 7 x 10 x 365, you also have to factor in the 24-48 hours where the clinical team is waiting on an answer. Given that last piece, the potential waste could be 25,000 hrs x 48 hrs. In other words, up to 1.2 million hours wasted per year that impact the efficiency of clinical trials.
🤯 According to that math, Genentech can, in theory, save 136 years per year. Yes, you read that right. Welcome to the power of automation.
To bring these numbers closer to home, Jim points out that some of Genentech’s studies involve treatments for COVID-19, with medications being sent to patients experiencing pneumonia-type symptoms at hospitals.
A matter of 15 minutes can mean a person’s life if they have trouble breathing. And 15 minutes is just one person answering one question, not even accounting for lag time because of time zones, so this all adds up. If a chatbot can get people instant answers, and we can ship a drug out to a patient faster, that impacts the patient’s life.
How will chatbots impact job responsibilities at Genentech?
According to Jim, the support piece is added on top of the tasks that employees typically handle. Automating some of the repetitive tasks will free up Genentech’s team to do higher-value work, such as responding to complex requests faster.
Moreover, the research groups can spend less time searching around and waiting for answers, and spend more time focusing on innovations that make the world better. Jim says:
If we can speed up trials, we can get drugs out to patients and improve people’s lives. It affects people, and these people go out and affect other people, so the effect is exponential.
Jim’s advice for enterprises looking to build a Google Cloud Dialogflow agent:
Make sure you understand what the most common questions are. Prioritize which problems and questions you want to attack conversationally. Think about how you want to serve up the responses and make them as brief and clear as possible.
While Jim and his team were busy cranking out the questions and responses according to priority, our team at Botcopy helped out by rewriting and programming the chatbot’s responses to be concise and digestible. Here’s a glimpse at a before-and-after.
(Genentech’s bots are internal-facing, so the words are obscured on-purpose for privacy.)
Regarding the edits we made, Jim said:
It made it a lot easier to read. I definitely enjoy working with you guys, and I look forward to working with you some more.
The feeling is mutual, Jim! One of the goals of launching Botcopy is to help make the world a better place. Genentech’s use case is the kind of thing that gets us out of bed in the morning.
As for Genentech’s chatbots, they are deployed and garnering positive feedback. Jim reported the following:
From end users up to upper-level management, the feedback has been great. It was mentioned at one of our town hall meetings that we had a bot that was part of their task force that came up in less than 24 hours, so they were super excited about it.
Can a chatbot help your business, too?
Yes, the right chatbot can help most enterprise firms save time, cut costs, and boost productivity. Significantly. And the best part is, having a smart chatbot isn’t only good for business; it’s also good for the world. Just imagine: no more boring wait-times for customers, and fewer tasks that require people to act like robots. 👍
If you need help setting up your Botcopy bot or Google Cloud Dialogflow agent, feel free to get in touch, or sign up free.
