What’s happening with AI in constituent engagement?
BY ANNE MEEKER
When ChatGPT entered the public imagination in December 2022, the implications for Congress’ interactions with constituents felt important — but unclear. Would chatbots revolutionize the way Congress and constituents communicate? Would GenAI lead to a flood of generated outreach that would overwhelm Congressional offices? Would malicious actors use voice cloning to deliberately target Congress?
Two years later, our team talked to staff members in Congressional offices who handle constituent input (including members of the wonderful Modernization Staff Association), along with entrepreneurs pitching new AI-generated advocacy tools, and with CRM (constituent/customer relationship management software) vendors providing the basic tools all Congressional offices use to handle mail. We asked each of these groups whether they have seen constituent engagement impacted by AI in the last few years, and how they imagine it will evolve in the future.
On constituent engagement, AI is the dog that hasn’t barked yet.
Despite GenAI’s clear potential to change how Members and constituents communicate at scale, its impact so far has been minimal. However, this is more a reflection of the ossified and outdated system of at-scale constituent engagement with Congress, rather than an indictment of these technologies themselves.
Here were the big takeaways:
The overall quantity of constituent correspondence is mostly unchanged.
To be more precise: offices have not measurably noticed a change in their workloads related to handling mail, email, and phone calls since the advent of ChatGPT. There are certainly some caveats: junior staffers who handle constituent input turn over frequently, so many do not have long-term insight into year-over-year changes. But on the whole, when asked whether they have seen any change from GenAI, most staff shrugged.
To explain briefly how this type of engagement-at-scale works: whether a constituent reaches out via phone, email, or mail, their “opinion” is logged in the Congressional office’s CRM. Alternatively, large-scale letter-writing campaigns from advocacy organizations can also send large batches of emails directly through the Communicating With Congress (CWC) API. From there, the office will “batch” multiple opinions on the same issue to receive specific form letters explaining the Member of Congress’ stance on the issue, or if the letter warrants, more personalized responses like phone calls or custom letters. All of the approved CRMs for Congress offer features that automatically batch some letters, but many Legislative Correspondents (LC) and Staff Assistants (SA) tasked with managing incoming mail still batch many messages by hand.
There may be a few reasons for this perceived lack of an impact:
The advocacy vendors using AI tools to help generate or enhance constituent opinions mostly use the CWC API, which makes it easier to batch in CRMs because it comes in with consistent metadata (campaign ID and issue tags) — so the area of constituent input most facilitated by GenAI was already the easiest one for offices to process.
Improvements in other features around mail-processing for Congressional offices may be cushioning any changes in overall volume: Member office websites usually ask constituents to select issue-area tags when submitting emails, which facilitates batching. Although there is clearly still some frustration and room for improvement in CRM auto-batching tools, they have also improved to the point where they can at least handle a “first pass” on general constituent input, reducing the overall workload on staff.
The one qualitatively new area is AI-generated phone calls designed to provoke an emotional response.
The one area we found where GenAI has enabled a new model of constituent engagement was the “Shotline” and its offshoots: in January and February of 2024, gun control groups Change the Ref and March For Our Lives created an online advocacy tool that allowed individuals to call Congressional offices using the recreated voices of children killed by gun violence. Houston Public Media reported that more than 160,000 calls were made using this tool. One Senate office at a Modernization Staff Association Roundtable noted that they had received similar generated calls purportedly from children killed in Gaza.
Offices described the impact of these calls as intense, but short-term: hearing repeated violent stories designed to be shocking was difficult for the interns and staff, especially when the tool first rolled out and offices were receiving hundreds of calls per day. Some offices turned off their phones entirely, sending constituents to voicemails that could be processed more slowly and with greater emotional distance. However, they quickly faded away, and have not represented a major percentage of call volume in a sustained way since.
As one perceptive Staff Assistant put it, in explaining why these tools have not been more widely used: part of what constituents want when they call their Member of Congress is the reaction of the person who answers the phone. Pressing a button to make a call offers no immediate feedback, or the ability to see the emotional response provoked by the call — meaning that there is less reason for people to continue repeating these calls.
The novelty of this approach is not necessarily the impersonation: individuals and activist organizations have given false constituent names and addresses, recruited actors, or even contracted with call centers to pretend to be constituents before. The novelty here is the deliberate goal of provoking a personal, visceral, emotional response for Congressional staff through the use of these tools — engagement in a punitive mode, facilitated by GenAI.
These calls also make up part of an emerging cycle of spikes and doldrums for Congressional offices on the phones: beyond the Shotline calls, every office we talked with brought up the flood of calls from constituents around Congressional efforts to ban or force the sale of TikTok, driven by a pop-up ad from TikTok itself. While Congressional offices have always received more calls around breaking news or high-profile events, this cycle seems to be more exaggerated than in previous years — partly due to the proliferation of AI tools like the Shotline, but also from social media algorithms (or companies themselves) sending outraged constituents to the phones more quickly than has previously been possible.
Offices are, for the most part, not using AI tools to draft constituent input.
It is also worth noting that most offices are not currently using GenAI tools to draft responses to constituents. Of the LCs and SAs we spoke to, none said that they actively use GenAI tools in their work, and that the pressure from management around AI was more geared toward preventing interns from using AI tools to complete assignments than increasing the pace or quality of output.
While some offices in Congress are experimenting with new technologies (and, per the House Chief Administrative Officer, GenAI tools are currently only approved for experimental purposes), others noted concerns with constituent trust, or the risks of accidentally incorporating false information into constituent-facing correspondence — especially for interns using these tools without experience and context.
Where does this leave Congress and constituents?
So where does this leave the ecosystem of constituent engagement? While advocacy vendors might be advertising GenAI tools that facilitate more or more unique constituent communication, these tools have not increased the effectiveness of constituent input within Congress, and Congressional offices have been able to cope with any increased workload through non-AI tools. These new technologies are — so far — nibbling around the edges of constituent engagement, not contributing to any major overhaul of the system.
And looking at the participating players and their incentives…maybe this is not surprising.
From a Congressional office perspective, the “point” of at-scale constituent engagement is to demonstrate responsiveness: to respond to a phone call, an email, a postcard, a petition, etc., by sending a form letter that shows the office is listening, and is an organic opportunity to share information on the Member’s positions and activities. At-scale input is only one small variable of many in directing policy, in part because offices understand that it is unreliable as an indicator of public sentiment. Offices cannot use quantitative data from their CRMs to accurately assess constituent sentiment, given that volume is so determined by outside advocacy groups and media coverage. It is also rare to receive input that is the magic combination of:
novel, actionable information,
directed to a Member on the right committee or leadership who is also senior enough to shape policy, and
timed to be useful in the process of developing a bill.
Offices universally emphasize how much they value hearing from their constituents, but in parsing through the avalanche of mail, the most directly useful information is local intelligence that can shape an outreach program.
None of this is a bad thing — responsiveness is certainly part of a functioning Congress — but it leaves constituents stuck in the middle. The mismatch between advocacy groups and activists urging people to call Congress as a method of impacting policy, and Congress’ inability to use that contact for any policy-related purpose, is a recipe for civic frustration.
But GenAI for constituent engagement is the dog that hasn’t barked…yet.
The lack of impact of new technologies in the current system should be a signal to both Congress and advocacy groups about the limited effectiveness of at-scale communications, and a challenge to rethink the nature of constituent engagement.
Constituents engaging with Congress at the behest of advocacy groups should ask critical questions about the expected impact of their engagement, and the organization’s theory of affecting the Congressional agenda.
Advocates wondering why at-scale engagement is not delivering results should consider engaging with ongoing efforts to supplement Congressional capacity.
And, lastly, Congress should think critically about where, how, and what types of input from constituents would actually be useful in policymaking, and be empowered to experiment with new modes of engagement that solicit useful information. Responsiveness via form letter is only one part of the complex equation of making constituents feel heard and rebuilding trust in the legislative process.
Empowering offices to experiment requires some risk tolerance from the institution and from individual offices — but offices must weigh this risk against the opportunity costs of continuing to staff at-scale communications instead of exploring new modes of engagement that might be more productive, rewarding, and resilient for the long term.
