Humach is a company that combines the strengths of both humans and machines to deliver exceptional customer experiences. Pulse 2.0 interviewed Humach CEO Tim Houlne to learn more about the company.
Tim Houlne’s Background
Could you tell me more about your background? Houlne said:
“Sure. I started in banking and finance and got into cellular in its infancy stage, ending up running a team of 40 people providing customer service and sales, but you did not call it a contact center back then. Went to work for one of the leading BPO companies and then in the late nineties, realized technology had advanced enough to allow for remote working, so I became CEO of Working Solutions, one of the original pioneers of the work at home model. Then in 2012, I wrote The New World of Work, From the Cube to the Cloud, suggesting that businesses would shift from bricks and mortar to work @ home, where employees could work remotely. This was long before the pandemic when the entire world worked from home in 30 days.”
“I now think we’re at another pivotal moment in history where work is about to change again. As outlined in my latest book, The Intelligent Workforce, How Humans + Machines Will Co-Create a Better Future, we live in a world where everything is digital, and the lines between humans and machines continue to blur at an unfathomable rate. Businesses that continue to invest in AI will accelerate and widen the gap between those that do not.”
Formation Of Humach
How did the idea for Humach come together? Houlne shared:
“It is hard to believe, but 10 years ago, I started to realize that AI will change our working and professional lives. Since my background was customer experience and CX is one of the top use cases for AI, I started Humach (humans + machines) to focus on AI and automation. The benefits of AI for CX are undeniable, creating a superior customer experience, more efficiently with less resources.”
“Our flagship AI solution for enterprises is mAI Pilot, an AI / CX solution in a box that lets organizations rapidly build and train custom language models based on the corpus of data under their roof. This could be product documentation, call center logs and recordings, video tutorials, etc.”
“We also create custom language models (CLMs) for mAI Pilot, that are more practical and accurate for enterprises because inference results are based on their data, not on the whole of the Internet, which is how most major LLMs are currently trained.”
“In 2023, we added an AI Certification Training program for our agents, upskilling them to prepare and thrive in the new world of AI. We call them ‘AI Whisperers.’”
Onboarding AI Solutions
When business and IT leaders are thinking about implementing AI, it seems like a daunting task. How should businesses be thinking about onboarding AI solutions? Houlne noted:
“This may seem counterintuitive, and it is where I see companies stall out in their AI efforts. Since AI is so new, they try to evaluate solutions in much the same way they’d evaluate a CRM, ERP, or CCaaS platform or workplace productivity suite. But these are mature solutions spaces where AI is nascent. One of the myths is that IT needs to drive the AI strategy, but we see business owners now able to drive their own strategy with less internal IT resources required for AI deployments.”
“What I recommend is that enterprises flip the script from thinking about major purchases and implementation efforts to much smaller, bite sized projects. Take the call center for one unit, for example, and not even your biggest one. Bring in a pilot and learn and include both internal and external use cases where AI can assist. Maybe this is the right solution, and maybe it isn’t, but you won’t spin cycles trying for a one-size-fits-all solution. If the pilot is a success, you can expand from there, knowing that you’ve de-risked the project. This also helps sell it across the enterprise.”
Avoiding Having AI Give Customers Bad Info
How can businesses avoid having AI give customers bad information? Not just hallucinations, but answers that may be correct broadly speaking but which are not germane to the enterprise itself, e.g., in healthcare or financial services. Houlne replied:
“This is critical for all AI deployments. Data integrity is key. Bad data = bad results. Regardless of where companies are on their AI journey, improving data will be required for successful AI deployments.”
“Businesses should really be thinking about this in two ways. First, consider deploying a custom language model (CLM) instead of a mass market large language model (LLM) that was trained on the whole web. You’ll get targeted and relevant results, instead of broader, and potentially irrelevant results.”
“Secondly, you can lean on employees to help fine tune and provide ethical oversight for your AI instance. These may be people at risk of being displaced by AI, but with the right training, like Humach’s AI Tuner Certification program, frontline workers can be upskilled on AI, including how to improve and update results and how to make sure bias and other factors are not limiting the results provided.”
Measure Effectiveness Of AI Implementation
How can enterprises measure the effectiveness of their AI implementation? Houlne pointed out:
“Enterprises can measure the effectiveness of AI solutions in their call centers by tracking key performance indicators (KPIs) such as customer satisfaction scores, average response time, first-call resolution rates, and call deflection rates.”
“They can also analyze the accuracy and efficiency of AI-driven responses, evaluating whether the AI is resolving issues successfully without human intervention. Additionally, monitoring the reduction in operational costs, improvements in agent productivity, and the ability of AI to handle peak volumes can provide insight into its effectiveness. Gathering feedback from both customers and employees regarding their experience with AI can further help assess its impact on service quality.”
AI Becoming More Personalized
How can AI become more personalized? Houlne explained:
“One of the problems companies face is implementing mass market solutions from Big Tech. The LLMs were trained using wide swaths of the public internet and so therefore its no surprise that inference can not provide domain relevant responses.”
“Instead, companies should go small, meaning adopt a custom approach with a CLM. CLMs are only trained on that institution’s corpus of data and so the results are specific and highly targeted.”
“One of the biggest value propositions for AI is hyper-personalization. Interactions based on profile, prior history, and predictive analytics; this should be a major part of your AI strategy.”
Downside Of AI For Workers
With all the promise of AI, is there a downside in terms of the impact on workers, particularly frontline workers. What can businesses do to help? Houlne emphasized:
“There’s a big fear that workers will be displaced by AI. As we’ve seen in past technology waves, there is disruption, but employees who adapt can benefit. Take a customer call center with a new AI implementation: Those workers instead of getting displaced can be upskilled to actually make the AI instance more successful by using their domain expertise about the company, solutions and the customer to tune the model. There are AI certifications, including from Humach, that companies can use to upskill their workforce. In this way, the AI gets better, which benefits the enterprise, and the workers, instead of being obsolete, can participate in the new AI economy.”
“My favorite quote is that AI will not take your job, but someone who knows how to use / leverage AI will take your job. Careers will change, but history has demonstrated when technology displaces jobs, it creates more jobs in the long run.”
Customer Success Stories
Can you share any specific customer success stories? Houlne highlighted:
“Humach was selected by a major theme park and entertainment company to automate and improve its customer experience. Facing seasonal surges in customer calls and the resulting increase in hold times, the company wanted to improve CX and provide self-service solutions on a limited budget. Humach developed a custom language model (CLM) based on the company’s corpus of data, publicly available information, FAQ, pricing and operating hours.”
“Trained on this CLM, the Humach AI-based digital agent could troubleshoot user issues, provide guidance for self-service solutions, and assist site visitors. Generating 125,000 net-new conversations, Humach’s digital agent handled 42 percent of incoming chats, identified thousands of leads, and reduced headcount by 52 FTE. During the first year of operation, there was a 700 percent increase in intents handled by the digital agent. The typical time it takes for Humach to build an instance like this, including voice and chat capabilities, is between four and six weeks.”
Total Addressable Market
What total addressable market (TAM) size is the company pursuing? Houlne assessed:
“The total addressable market for AI software, including agentic AI, is projected to reach $4 trillion, according to Battery Ventures.”
Future Company Goals
What are some of the company’s future goals? Houlne commented:
“ Taking AI in CX mainstream. Today 30 percent of workforce tasks can be automated and yet major enterprises are wringing their hands about how to implement AI. Tomorrow, these same enterprises—assuming they don’t become obsolete through inaction—will be spinning up digital agents as part of their workforces. They will be upskilling employees to improve those AI agents. Humach is already helping companies do these things, and we’ll be in the middle of the storm tomorrow.”
Additional Thoughts
Any other topics you would like to discuss? Houlne concluded:
“There’s an interesting phenomenon taking place right now: Despite all the hype around AI, mainly advances in technology and discussions about American competitiveness, major enterprises are slow in adoption. Part of the choke point is that AI has become a C-Suite issue because it’s not just IT. It’s not just supply chain or CX. It’s everything. Either companies will figure this out, or they risk falling behind. The smart companies already have piloted, failed, learned and are moving forward with new initiatives. If I were the CEO of a major enterprise today who doesn’t have four or five AI pilots in flight today, I would be worried about my job.”
