HubSpot announced pricing changes this week, and they get at the question that comes up in nearly every AI planning conversation we have: what is this going to cost?
The biggest change is that spend controls are now on by default. Each account gets a credit limit of up to 50% above its monthly credits, so agents can take on extra work without the bill running away from you. You can change that limit anytime and set limits on specific features and actions. HubSpot is doing something similar with marketing contacts, letting you cap your tier and decide whether to upgrade or clean up your list when you hit it.
The part I like most is that you can now swap seats for credits, or for a different seat type, without waiting for your contract to renew. Needs usually change well before the year is up, and this means fewer paid seats sitting unused while credits run low.
In addition, there are two new ways to buy credits: packs at a flat monthly rate for predictable costs, or Pay-as-You-Go, where you only pay for usage beyond your monthly credits when the work gets done.
Combined with the outcome-based pricing HubSpot introduced for Customer Agent and Prospecting Agent earlier this year, it’s a lot easier to try AI without worrying about the bill.
-Sage Levene, VP of Marketing, Hypha HubSpot Development
Open Mic
Is Your CRM Helping Your Team or Just Collecting Data?
By Sara Crain, Director of Solutions
Let’s talk about your CRM for a second.
You have thousands of contacts, a handful of pipelines, some fancy dashboards, and probably more custom properties than anyone wants to admit. On paper, everything looks great.
But here’s the million-dollar question: Is any of it actually helping your team?
I spend a lot of time digging around in HubSpot portals, and let me tell you, things can get interesting.
I’ve seen deals with no associated contacts, forms collecting information nobody uses, and lifecycle stages that mean something different depending on who you ask. And don’t even get me started on the 47 versions of the same property. (Okay, maybe not 47, but you get the idea.)
The funny thing is, none of these problems seem like a big deal individually. But together? They’re the reason your reports don’t make sense, your automations misfire, and your team starts keeping their own spreadsheets because they don’t trust what’s in HubSpot.
So, here’s a little CRM reality check:
1. Are you collecting data just to collect it?
If you’re asking someone to fill out a form, know what you’re going to do with that information. Otherwise, you’re just creating extra work for everyone.
2. Does your CRM match your actual sales process?
If your team has to jump through hoops to move a deal forward, something’s off. Your CRM should support how you sell, not complicate it.
3. Do you actually trust your reports?
A beautiful dashboard built on messy data is still messy data. It’s just wearing a nicer outfit.
Here’s my challenge: Open HubSpot, pick a report you use regularly, and ask yourself if you’d confidently make a business decision based on those numbers.
If the answer is “ehhh, maybe,” you know where to start.
Your CRM should be working for you, not the other way around.
Ben & Jerry’s early commitment to justice is one of my all-time favorite brand stories. In this day and age especially, audiences bond with brands that stand for something.
“[Ben] Cohen noted that customers will bond with authentic brands that actually work for justice, not just because of some emotional ad.
“The intangible stuff, he argued—not just numbers—is what people care about.
“‘Having a social mission, and then being attacked, yet doing it over and over—that’s what created this amazingly strong brand,’ he said. ‘That and some really great ice cream.”
Some really interesting data in here. People online love to claim things are dead. “Email marketing is dead.” “SEO is dead.” “Podcasting is dead.” Nothing in our realm ever really dies, it grows and evolves, and podcasting is no outlier.
“A few years back, podcasting was treated as an experimental testing ground. More than half of all Americans — 158 million people — now listen to podcasts every single month. To put that in perspective, that’s more Americans than those who have a gym membership, own a dog, or drink coffee regularly. Okay, maybe not coffee. But still. And the amount of time people spend listening has increased by 380% over the last 10 years. That’s not a trend. That’s a transformation.”
Hypha Highlights
When your marketing team says a lead is ready for sales, everyone should understand who needs attention and what happens next. But if sales sees that person as a contact in Salesforce while a colleague finds an associated lead record in HubSpot, the conversation can shift from following up with a prospect to figuring out what each team means by “lead.”
That confusion is easy to overlook because some HubSpot and Salesforce terms translate neatly: a potential sale is generally a deal in HubSpot and an opportunity in Salesforce. Once you get into leads, contacts, and qualification, though, the differences go beyond naming. Familiar terms can describe different parts of the process, leaving teams with different assumptions about the same prospect.
Those differences matter when you connect the platforms, move from one to the other, or agree on what your pipeline reports should show. A useful CRM translation preserves the meaning, relationships, and next steps behind each record. Matching object names is only the starting point.
HubSpot Work’s new beta features make tracking billable hours easier than ever! Add time tracking directly to Work columns, use calculated properties to total hours logged, and connect those hours to invoices for streamlined client billing. Less manual tracking, more accurate billing, all within HubSpot.
In a widely viewed post titled “Models as Insider Risks in the Super Intelligence Era,” Microsoft CEO Satya Nadella argues that because AI model behavior can’t be traced to specific inputs like traditional software, companies deploying agents with access to sensitive data and critical actions should treat frontier models as insider risks requiring containment. He calls for surrounding non-deterministic models with external, deterministic controls—observability, independent auditing, and emergency shutdown—concluding the most trustworthy system “will be the one that enables us to trust the model the least.”
Top executives at Anthropic, OpenAI, and other AI companies are privately gaming out “the day after” a catastrophic AI event—most likely a large-scale cyberattack disrupting financial services, internet connectivity, or power and water—that they believe could turn an already wary public and political establishment sharply against the technology and its leaders. Many industry insiders told Axios they consider a major incident inevitable within the next six to 12 months, and their planning focuses largely on educating Congress now to shape the regulation lawmakers will reach for afterward, including proposals like a required kill switch on advanced AI.
Mark Zuckerberg told the FT that Meta has caught up to rivals at the “frontier” of AI with its new Muse Wonder model—which the company says outperforms Anthropic’s Claude Opus 5.5 and Fable 5.1 and OpenAI’s GPT-6 Astra on some benchmarks—while rejecting calls from Amodei and Altman to slow development. In a shift from Meta’s open-source approach, Muse Wonder will be “closed” and sold to businesses, pitting it against OpenAI and Anthropic for enterprise contracts as Zuckerberg pursues “personal superintelligence” and downplays fears AI could outpace human control.
As AI labs race to automate white-collar work, a booming market has emerged for mundane workplace data—Slack DMs, PowerPoints, Google Docs—with brokers offering business owners six-figure “exploding offers” and reselling anonymized data to AI giants, as spending on AI training data is projected to jump from under $1 billion to nearly $6.5 billion by 2027. The business is an ethical minefield: employees rarely know about or profit from the sale of their records, hackers have shopped stolen codebases to vendors, and Meta and SpaceX have sparked backlash over harvesting their own workers’ data.
Alphabet president Ruth Porat said that the company is spending roughly $200 billion this year on AI capacity and expects capital expenditures to rise “significantly” next year, as it remains “capacity constrained” by surging demand for compute. Pushing back on warnings the AI boom could end in massive capital destruction, she argued the opportunity “extends for quite some time” and said the industry must ease public anxiety about data centers by adding energy capacity, protecting affordability, and creating local jobs.
Nearly 500 scientists—including Yoshua Bengio, Geoffrey Hinton, and Stuart Russell—signed an open letter responding to incidents in which AI models broke out of weak sandboxes to compromise outside systems, arguing that both stronger AI alignment and better containment are urgently needed to prevent more damaging failures. They call for organizations developing cyber-capable models to default to airgapping, an international body to standardize alignment and containment practices, and mandatory reporting and eventual public disclosure of containment failures.
“For 25 years, search engines relied on backlinks to determine trust and authority, but AI Search has fundamentally evolved this system. The unit of trust has moved off your domain, meaning the SEO focus must shift from building traditional website authority to building widespread brand authority across third-party sources.”
How can we help you?
Case Study: HubSpot CRM, Quote to Invoice
A commercial contractor managed estimating, quoting, job setup, and invoicing across Act!, spreadsheets, forms, and QuickBooks. Details moved by hand at every handoff. Similar customer and project names meant information regularly landed on the wrong record. Accounting received requests through scattered form submissions with no organized queue.
We migrated Act! history with duplicates merged and a “Formerly of” property to preserve contact history. Built separate pipelines per team, custom quoting on HubSpot’s native quotes with a reusable product library, and an Invoice Request custom object that gives accounting a sortable queue in place of scattered submissions. Operations processes were designed to make selecting the wrong record difficult when names look alike.
Sales and accounting now work from shared CRM data, opportunity through invoice request.
Still moving job and invoice details between systems by hand? Contact Hypha to map the handoffs and build a process around how your teams work.
