Beneficial State Bank is in a bind. As a certified B Corporation, the Oakland, California-based community development bank is built around social justice, climate accountability, and financial success without funding fossil fuels or private prisons. Artificial intelligence could help it meet those goals and make lending more inclusive. But using it requires some ethical gymnastics.
Until now, Terra Neilson, Beneficial State Bank’s chief impact officer, says it was easy to put AI in the “unethical” category. Its environmental footprint and social harms are regular headlines. But a recent project complicated that view. Neilson says the bank just extended a pilot run by Beneficial State Foundation, its nonprofit majority shareholder, using Stratyfy’s AI-assisted credit decisioning tool to improve efficiency and reduce systemic bias in credit underwriting.
BetterFi, a Tennessee-based community development financial institution participating in the pilot, has already used the tool to increase approvals within BIPOC communities by 21 percent. Neilson says if the banking system can use AI to find and correct those blind spots, it could help reduce the redlining and bias that have long kept marginalized groups from accessing capital.
Doing so, however, means weighing AI’s benefits against its costs. “You get decision fatigue,” Neilson says. “We’re being really intentional in this moment to say, what are the harms and what are the benefits that are worth the costs that we see?”
She says being a B Corp and mission-driven organization requires scrutinizing technology and vendor relationships to make sure they align with the bank’s intent. “AI is definitely in a world of its own because of the potential to bury the trail between decision and impact,” Neilson says.
The bank is part of a growing group of B Corps trying to reconcile AI with their environmental, social, and governance commitments. As backlash grows over AI’s carbon footprint, which recent research equates to New York City’s entire carbon footprint last year, B Corps face both an accountability challenge and a reputational one.
New world, new standard
B Corp is one of the world’s most rigorous and recognized environmental, social, and governance certifications, with more than 10,000 member businesses across 160 industries and 105 countries. Last year, B Lab, the global nonprofit behind the designation, released new performance standards across seven impact topics, including climate action, human rights, environmental stewardship, and justice, equity, diversity, and inclusion. The standards are more stringent than past rules and took about four years to develop, a period that overlapped with the rapid adoption of AI.
Clay Brown, B Lab’s chief standards officer, says the standards are equipped to handle the technology and its impact. “AI utilization is effectively a stakeholder governance question,” he says. “Whether it’s AI or choosing a supply chain that imports from across the world, how does that choice impact the stakeholders that your business is engaging with?”
Brown points to data centers as one example. A B Corp that uses AI in its workflows or products needs to consider where its data comes from, where the large language models powering the tool are hosted, and what impact those systems have. “These data centers, do they have any environmental standards that they’re following?” he asks. “What is the environmental impact of that? Or what about the communities that were either displaced or struggling with those data centers?”
Brown says the standards also call for B Corps using AI for content generation to develop ethical guidelines, while companies with carbon reduction plans are expected to account for AI usage in their calculations.
“We weren’t prescriptive… It’s a fast-moving space, and AI can be really useful,” Brown says. “We don’t want to say you can or can’t use it; what we want to say is you need to understand the impacts of using it and take steps to mitigate those impacts.”
Still, there are major gaps in the data, says Alex de Vries-Gao, a researcher at VU Amsterdam who has spent the past decade measuring the carbon footprint of emerging technologies. More recently, he has examined the water consumption and carbon footprint of data centers used by Microsoft, Apple, Google, and Meta to run AI models. “The suppliers are not transparent,” de Vries-Gao says. “How are you going to be properly transparent and accountable if you are not getting the info that’s required?”
For B Corps, that creates a practical dilemma. Do they use tools like Microsoft Copilot to generate content, write pitch decks, or brainstorm ideas while accepting that they may not know the exact environmental cost? Or do they avoid AI tools and risk falling behind competitors that use them? “This is ultimately the trade-off that you need to make,” he says.
Preparing for the backlash
Mark Lowe, founding partner at Third City, a London-based communications firm and certified B Corp that works with major B Corps, including Nespresso, says boycotting AI is not realistic. But he expects more scrutiny of AI’s environmental footprint.
“It is something that people will have to think about,” Lowe says. Third City is making decisions about AI and vendors in real time. A few months ago, the firm dropped its OpenAI subscription and switched to Claude over Anthropic’s stance against how the U.S. Department of Defense wanted to use its technology.
But the choices are rarely simple. Earlier this year, Third City launched GEOView, a generative engine optimization tool that helps marketing teams understand how their brands appear on platforms such as ChatGPT, Gemini, Perplexity, and Claude. The tool requires multiple searches on LLMs, which are already performing multiple searches themselves, Lowe says. “Obviously, huge amounts of searches mean a large environmental impact.”
Third City also uses Ecolytics’ Offset AI, which monitors and calculates the carbon footprint and water consumption of leading generative AI tools, then automatically purchases carbon and water offset credits. “Obviously, offsetting has a mixed reputation amongst the B community,” Lowe says. “But you’ve got to do something.”
Ecolytics’ main business is helping B Corps collect and measure the data needed for certification. Its platform already lets users calculate emissions from AI usage. Hazel Horvath, the company’s CEO and founder, says the Offset AI widget puts that impact in front of individuals in real time, narrowing the gap between values and action.
“We partner with B Corps and highly vetted carbon offset providers and actually have selected projects that are somewhat related to the AI supply chain,” she says. That includes water restoration, energy efficiency, and habitat restoration in areas where data centers are located.
Prompting responsibly
Yulu, a Vancouver-based public relations firm and B Corp, has taken a different approach by trying to reduce its footprint through better prompting. Melissa Orozco, Yulu’s CEO and chief impact officer, says the company uses Slack to workshop larger and more complex prompts before sending them along to an AI tool. “That’s a huge energy- and water-saving step that nobody takes,” she says. “People just go back and forth with AI and have a conversation… You should have a conversation with your team first.”
The company has an Ethical AI policy that lays out how and when employees should use chatbots. “We don’t ever talk to an uninformed agent,” she says. Yulu’s agent is named Rosie, a Jetsons reference, and is designed for unbiased, equity-focused prompting. “She’s got all our knowledge base, our company policy, our AI policy, our brand guidelines, all of this stuff so that we don’t have to continue going back and forth,” Orozco says.
Employees are not expected to memorize the guidelines. Instead, they are built into a universal ethical AI prompt that applies an intersectional equity lens and language check to the agency’s custom, client-specific AI engines. Ethical compliance is the default, Orozco says. The Yulu CEO admits it’s still a work in progress.
Yulu aims to update its guidelines every six months, and its relationship with AI keeps changing. Orozco says team members frequently express concern about the environmental cost and compute footprint of generative AI tools. “They are highly conscious of the emissions tax of generative AI,” she says.
That may be what makes B Corps well positioned to shape how the wider business culture balances AI’s benefits with its environmental costs. “We’re fine with putting in the work,” says Orozco. “We are not programmed to do things the easy way.”
