Emily Murphy, Administrator Of The GSA Shares Her Thoughts On AI
There is nothing ordinary about the year 2020, and in this highly charged political year, everything gets more attention than might have been deserved in previous years. A few weeks ago, Emily Murphy, Administrator of the General Services Administration (GSA) started making waves in the news. For many who may not have known her before, she has been leading the GSA for a few years helping bring innovative programs and initiatives to the agency.
Among many things that Emily Murphy is responsible for, one of the most consequential every four years is the ascertainment of a Presidential election winner so that a transition of power can proceed. That particular aspect of the GSA got a lot more news and publicity over the past few weeks than perhaps it ever has in recent history. Even those that deal with the government regularly might not have been aware of such a pivotal role that the GSA has with regards to elections.
Indeed, there are many other things that the GSA is much more known for being responsible for, from its role as the primary manager of real estate and buildings for the government to procuring billions of dollars of products, goods, and services. It’s in that latter light that I had the opportunity to interview Emily on how the GSA is approaching artificial intelligence. Automation and AI have served a particularly important role at the GSA, impacting how it goes about running its operations and procuring solutions on behalf of government agencies.
In a recent AI Today podcast, recorded right before the election, Emily Murphy shared her insights into how AI is transforming the federal government. In this article she further shares her insights into AI, the GSA, and the federal government.
During your time at the GSA you helped to launch the Centers of Excellence program. Can you share what the CoE is and how it’s helping advance use of data?
Emily Murphy: I was a senior advisor at GSA, prior to being confirmed as Administrator, and one of the areas I focused on was how GSA could better manage the intersection between contracting and technology innovation. GSA’s Technology Transformation Services, which is part of our Federal Acquisition Service, worked with other agencies to launch the first Centers of Excellence in late 2017, with the first partner (also known as our lighthouse agency) being USDA. We now have announced ten partnerships with different agencies where we have a presence. The Centers of Excellence teams provide technical expertise in support of the following areas: cloud adoption, contact center, customer experience, data analytics, infrastructure optimization, and artificial intelligence.
I am especially excited about our AI CoE, which is the sixth and latest pillar in our Centers of Excellence program. With the AI CoE, TTS brings together machine learning, neural networks, intelligent process design and Robotic Process Automation (RPA) to develop AI solutions that address unique business challenges agency-wide. The team provides strategic tools and infrastructure support to rapidly discover use cases, identify applicable artificial intelligence methods, and deploy scalable solutions across the enterprise.
Another highlight has been our partnership with the Joint Artificial Intelligence Center at the Department of Defense. We recently marked our 1 year anniversary of the JAIC and CoE partnership and released an announcement of our many achievements. One of the first things CoE worked on was assisting in the creation of the First Five Consortium, a first-of-its-kind Public Private Partnership which seeks to augment Humanitarian Assistance / Disaster Response efforts with AI-enabled capabilities. The objective is to provide no-cost AI-enhanced imagery to incident commanders and first responders within 5 minutes of a disaster to rapidly accelerate coordination times, saving lives and property. The CoE supported JAIC in developing a robust AI Capability Maturity Model and AI Workforce Model to help gauge operational maturity across multiple functional areas.
What do you see as some of the unique opportunities the public sector has around AI?
Emily Murphy: There are a world of possibilities when it comes to the benefits of AI for the public sector. One core benefit of AI is that it allows the government to test concepts before spending money building them out. So, instead of having to build something to see if it works, we can now use computers to do the first level of testing virtually. AI, and specifically natural language processing (NLP), can be leveraged to streamline many processes that were previously manual or document driven.
AI is also being leveraged by the federal government as part of a strategy to understand specific areas of need, such as rulemaking and regulatory reform. A few examples: currently we’re using AI to analyze comments made during the “public comment period” of rulemaking, to update regulations so that they reflect today’s technology and products, to identify areas of overlap and duplication in the Federal code and in regulatory guidance and make it easier to streamline regulations, and to make predictions about the effect of regulations on stakeholders. AI can also be used to accelerate hiring and the onboarding process at federal agencies.
We’re focused on finding and creating smart systems, services, and solutions that make it easier to interact with the government on every level.
What do you see as some of the unique challenges the public sector has around AI?
Emily Murphy: A few of the challenges that the public sector faces when it comes to AI are: data cleanliness, managing data, and hiring tech savvy AI federal system owners to properly operate, manage, and evaluate the system.
There are also unresolved issues surrounding the responsible and ethical use of AI. These are hard problems and require thoughtful consideration and cooperation to come up with a solution. Responsibility and ethics are embedded in every stage of the AI development lifecycle from design to development to deployment, and must include continuous monitoring and feedback collection to ensure it is behaving as intended without causing harm or causing unintended consequences. It is not just a checklist you run through after a solution is developed.
Because the responsible use of AI occurs at every stage of the development lifecycle, we need to reframe what it means to use AI responsibly. People often speak of evaluating an AI system after it has been developed. We need to move away from that to a mindset where the use of AI must be thoughtfully considered at every step. And this means that it’s not just the job of an AI ethicist to ensure this. The technical developers of AI are making ethical choices as they are building the system, so they need to understand how those technical decisions are also choices being made from an ethical perspective.
Partnership with private industry will be critical to ensure we are building responsible AI solutions. Government cannot be buying a black box of technologies with no insight, explanation or oversight as to how it is operating. Government experts need to partner with industry as they build AI solutions to embed responsibility throughout. Monitoring, evaluation and updating models must be at the forefront of the process, not an afterthought after a solution is built.
Teams need to engage across the organization to establish oversight and audit procedures to ensure that AI and automation continue to perform as intended.
What are some of the ways GSA is currently leveraging AI and Machine Learning?
Emily Murphy: We have implemented automation, AI and Machine Learning in a variety of ways. Robotics Process Automation (RPA) is an automated scripting technique that is sometimes categorized in the same overarching ecosystem of technologies as AI. We have implemented an enterprise platform for implementing RPA and have automated many processes. We continue to see the vendor community enhancing software offerings with new capabilities that are powered by AI and ML in areas such as anomaly detection, natural language processing, and image recognition. We see great value in using those advanced capabilities in the tools we already use or in new implementations. We also are growing our data science capabilities to use predictive analytics as an extension of our existing analytics and data management capabilities. We are accomplishing this through both investment in our staff’s learning, as well as providing them with data science tools.
What are some ways GSA is hoping to leverage AI and Machine Learning in the next few years?
Emily Murphy: In general, GSA is looking to implement AI and Machine Learning technology provided by vendors for existing software, as well as implementing custom solutions using this technology. Here are a few examples of how we hope to leverage AI and Machine Learning:
- Using Natural Language Processing (NLP) to improve the speed and accuracy of document-review tasks, such as End User License Agreement (EULA) or solicitations;
- Performing predictive maintenance on equipment maintained by GSA;
- Performing sentiment analysis and categorization on comments and service tickets; and
- Using chatbots to answer questions from technology users.
How is the GSA engaging industry and private sector in your AI efforts?
Emily Murphy: The federal government is using crowdsourcing in dynamic ways to engage industry and subject matter experts across the USA to advance innovation with artificial intelligence. In fact, in just the past six months, challenge.gov has hosted over a dozen federally sponsored prize competitions that focus on the use of artificial intelligence.
This past summer, GSA hosted the AI/ML End User License Agreement (EULA) challenge which showed how industry could provide IT solutions by leveraging AI and ML capabilities within the acquisition business process. This was hosted on challenge.gov and received 20 AI and ML solutions from solvers.
Another exciting example is Polaris, for which we just issued a Request for Information. Polaris is a next generation contract worth $50 billion geared at small innovative companies.
What is the GSA doing to develop an AI ready workforce?
Emily Murphy: We launched an AI Community of Practice to get smart people from across government talking and sharing best practices, then we set up an AI Center of Excellence to put their knowledge to work. This is how we lay the intellectual infrastructure needed to support the tens of thousands of federal workers, contractors, and citizens who will be working with this technology.
GSA is also very interested in looking at ways to build up data science and AI skill sets across federal agencies, as well as engaging externally to attract additional data science and AI talent into government. The TTS CoE and AI Portfolio have hosted three webinars focused on AI Acquisitions. These webinars have discussed topics such as defining your problem for AI acquisitions, how to draft objectives for your AI acquisition problems, and understanding data rights and IP clauses for AI acquisitions for federal employees.
As well, GSA OCTO hosts a speaker series called Tech Talks for GSA employees. These are sessions designed to introduce and explain new and emerging technologies to the staff of GSA. Related tech talks have included AI/ML Overview and “Battle of the Bots.” Our Chief Data Officer’s organization operates a training program for GSA employees called the Data Science Practioner’s Training Program. This program develops the core data science skills underlying effective implementation of AI and ML.
How important is AI to the GSA’s vision of the future?
Emily Murphy: AI is a critical part of GSA’s vision for the future, and it should be for all agencies. Advances in AI and ML have fundamentally changed the way private industry does business. Government should be leveraging AI technologies in a responsible manner to serve the people of this country. GSA specifically will better be able to support our partner agencies through faster, more efficient and more informed mechanisms with the support of AI.
How is AI helping with GSA’s mission today?
Emily Murphy: Like many private companies, we have lots of work – more than our workforce can always handle. We have also discovered that many of our current processes exist simply because we have always done things that way. As well, when we would get new systems, we’d program them to automate the old process without thinking through whether there was any value in that and whether the process was one we wanted or needed. AI is allowing us to modernize our systems and processes, and shift from low value work to high value work. We have over 70 bots currently in operation that have saved over 260,000 hours. Here are a few examples:
- Made in America Robomod – This bot goes through and validates the more than 60 million items on GSA schedules by examining supply chain risks and confirming where items are built. This bot saves countless hours on rote work.
- Truman 2.0 bot – prepopulates price negotiation models with other information we have
- A bot was used to bilaterally modify GSA’s leases in response to a new law that bans the use of certain technologies in those spaces in order to improve the security of federal employees and their data. When you keep in mind that GSA leases about 180 million square feet of office space in addition to our owned space, this would have been a herculean task without RPA.
- GSA recently created a guide on how other agencies can use GSA contracts to access AI in the rulemaking and regulatory process. AI can help regulators reduce duplicative/outdated regulations, understand burden and impact, and automate time consuming manual research and analysis tasks. .
- GSA is promoting Communities of Practice as well to help other agencies as they adopt AI, ML, and RPA.
What AI technologies are you most looking forward to in the coming years?
Emily Murphy: I’m looking forward to universal communication across languages to around the world, and the accelerated digitization and processing of government forms. Many AI technologies and tools hold promise of helping the federal government to become more efficient, reveal greater insight, and make better decisions. The key is to ensure that we leverage both human and AI systems together.