Interviewing father and son authors Richard and Daniel Susskind in April of 2016, NPR's On Point host Tom Ashbrook explored the thesis of their recent book, The Future of the Professions: How Technology Will Transform the Work of Human Experts (Susskind and Susskind ). Society, the Susskinds aver, is on the brink of fundamental and irreversible changes in the way that professional expertise is delivered. Specifically, “increasingly capable machines will transform the work of professionals, giving rise to new ways of sharing practical expertise in society” (Susskind and Susskind , 303). Current professions are an artifact, they say, “built to meet a particular set of needs [but] in their current form, will no longer be the best answer to those needs” (Susskind and Susskind , 3). Citing professions the Susskinds believe at risk of serious disruption from automation, Ashbrook notes: “People have talked about lawyers and the law for a while. You've got doctors right there; we think of those immediately. But you also tick off accountants, management consultants, architects, teachers, journalists and, heaven forfend, even clergy” (Ashbrook ). Yes, the Susskinds believe that not even clergy are immune from the disruptive technologies that are altering the landscape of work in the twenty‐first century.

Clergy, they argue, function largely as gatekeepers mediating access between believers and sacred texts or between worshipers and God, and traditionally operate within the context of localized religious communities. But current and emerging technologies provide alternative, nonlocalized means to deliver religious knowledge and analysis—Bible websites, streamed worship services, clergy blogs, digital‐only congregations, video pilgrimages, online religious scholarship, and so on—which could have a profound impact on the religious status quo.

These platforms and systems can strengthen people's existing religious beliefs and can also support entirely new internet‐based faiths. Sometimes, though, the services encourage people to question their inherited religions. Almost two‐fifths of young practicing Christians in the United States use the internet to fact‐check the statements that their religious leaders make (Susskind and Susskind , 66).

Just as Gutenberg's press democratized access to the Bible in the fifteenth century and broke the clerical monopoly on scriptural knowledge, the internet is today democratizing access to the spiritual realm, with a potentially devastating impact on the current order, the Susskinds aver.

Not everyone agrees. Legal scholar Frank Pasquale believes that the Susskinds’ claims are vastly overblown because “software has a difficult time mimicking the spontaneity, creativity, flexibility, and perceptiveness that are the hallmarks of good nurses, surgeons, psychiatrists, pediatricians” and, we assume, clergy. In a scathing rejoinder to the Susskinds, he mocks them as predicting “the emergence of bot preachers, ministering to avatars in Second Life's virtual cathedrals” (Pasquale ). (It's worth noting, however, that since 2007, a virtual “Anglican Cathedral” has existed in the Second Life virtual world, with daily worship, a weekly Bible‐study class, and counseling services.) Nevertheless, many tasks currently performed by nurses, surgeons, psychiatrists, and pediatricians are susceptible to automation. Surely, the same is true of clergy.

An extensive 2013 Oxford University study (Frey and Osborne ) evaluated 702 common occupations. Using a sophisticated ranking algorithm, each was scored for likelihood of automation. Many current jobs are in the crosshairs of computerization. The study concludes, for example, that telemarketers, loan officers, and cashiers are likely to see their jobs replaced by automation with above 95 percent likelihood. Clergy, on the other hand, had only a 0.83 percent chance of losing employment to automation according to the study. Seminary students, take heart!

So, is emerging technology likely to disrupt the profession of clergy in the way that it promises to disrupt other white‐collar professions? In a word, no. In the foreseeable future, churches would not be hiring/purchasing robot clergy to fill a vacant pulpit. But technology definitely will impact the way that clergy perform their functions, for good or ill, even in the short term. In the long run, who knows?

In this article, we examine the likely impact of digital technology, especially artificial intelligence (AI), on the profession of clergy. We identify two distinct technological trends that promise to disrupt the practice of working clergy in America. We will focus primarily on Christian clergy, but our conclusions differentially apply to non‐Christian religious leaders as well.

The first trend we call online spirituality. This is the delivery of religious “products”—worship services, pastoral care and counseling, inspirational content, sermons, reflections, prayers, scriptural scholarship, and retreats—via digital media rather than within traditional congregational settings. A cursory Google search shows that these are already widely available. The modes of delivery may be nontraditional, but the content is still produced by human beings, often clergy working within the context of traditional churches. Tellingly, nearly all of the Susskinds’ examples (Susskind and Susskind , 61–66) of current incursions of technology into the sphere of religion are of this ilk.

A second, more futuristic possibility looms: that artificially intelligent digital systems actually will supplement or supplant human clergy in significant ways— writing sermons, delivering pastoral care, conducting scriptural/theological research, or performing sacramental functions. We will refer to this possibility, perhaps a bit whimsically, as digital clergy. To date, instances are few and mostly one‐offs. For example, a robot has officiated at a Japanese wedding (Daily Mail ); another is available to perform funerals (Gibbs ). However, such technological incursions are rapidly disrupting other professions such as medicine, law, journalism, and finance.

Automated systems currently diagnose diseases, perform surgery, write wills, do legal research, compose news stories, and trade stocks. The fact that computer scientists and AI researchers have not as yet expended much effort automating many of the tasks traditionally performed by clergy does not mean that they could not do so or would not do so in the future, given sufficient incentives. How realistic then is the prospect that significant functions currently performed by human clergy can and will be performed by the “increasingly capable machines” heralded by the Susskinds?

The theological and pastoral issues raised by online spirituality have been discussed elsewhere (e.g., Campbell ). Some theologians also have discussed the issues entailed by AI (e.g., Herzfeld ). But the possibility of digital clergy raises some separate theological questions: To what extent can a digital system legitimately be said to minister? Does the ontological status of a machine comport with our notion of a sacramental agent? Can a machine be ordained? Assuming that digital clergy can exist, how important is it that they function in the same way as human clergy? Must a digital minister have genuine religious conviction (faith)? What would that even mean?

Our goal in this article is twofold. First, we will describe some of the ways in which technology is currently changing the practice of ministry and some ways in which it promises—some would say threatens—to do so. Second, we will point to some theological and practical issues that these changes raise. The article is arranged as follows. In the following section, we discuss briefly the current state of the art in artificially intelligent systems, especially those ushering in the “postprofessional society” predicted by the Susskinds. The section on “Online Spirituality” describes how current trends in digital religion are affecting the practice of religion today and the likely effect on clergy in the future. Then, in the section on “Digital Clergy,” based on trends in other professions, we describe some technologies that could automate key functions clergy perform today, specifically preaching and pastoral care. The section “Theological Issues” describes some theological issues raised by the use of digital systems to perform ministerial functions. In the final section, we draw some conclusions of this research.

THE AI REVOLUTION

AI promises to be one of the most disruptive technologies of the twenty‐first century. A recent study panel of experts on AI from academia and industry provides the following definition of AI: “Artificial Intelligence (AI) is a science and a set of computational technologies that are inspired by—but typically operate quite differently from—the ways people use their nervous systems and bodies to sense, learn, reason, and take action” (Stanford University ). Progress in AI has been uneven since computer pioneer John McCarthy coined the term AI in 1956. Periods of progress and promise, often accompanied by excessive hype, have been punctuated by the so‐called AI winters during which progress stalled and funding dried up. However, the past decade has seen accelerating progress. Consider the following:

• Video games powered by computer graphics, vision, and AI planning today constitute a larger entertainment industry than Hollywood.

• Automated systems routinely outcompete the best human experts at chess and even the popular TV show . For the first time, a computer program has become competitive at world‐champion level Go, one of the most strategic of board games.

• Google's Translate utility now translates among 140 natural languages, rivaling skilled human translators (Lewis‐Kraus ).

• A recent article in the Smithsonian's online magazine lists nine surprising tasks that robots are currently performing, including cooking dinners, filling prescriptions, checking guests into hotels, training athletes, and even riding camels (Matchar ).

Such advances in machine proficiency largely have been driven by two major factors: Moore's Law and advanced sophistication in machine learning (ML) algorithms.

Moore's Law describes exponential growth in the production of semiconductor electronics. Originally proposed by Intel founder Gordon Moore in 1965, Moore's Law predicted that component density (transistors per square unit) on integrated circuits would double approximately every 1.5–2 years, a trend that has now prevailed for over four decades, though futurist and inventor Ray Kurzweil has noted (Kurzweil ) that computers were actually doubling in power long before transistors were invented.

Accompanied by advances in storage capacity, materials science, fiber optic transmission, and multicore architectures, the result of this technological progress has been a remarkable explosion in the computational prowess of digital systems. In introducing Moore at a 2015 event, Intel CEO Brian Krzanich observed that, compared to Intel's first generation 1971 microchip, Intel's latest chip offered 3,500 times more performance, was 90,000 times more energy‐efficient, and some 60,000 times cheaper (Friedman ). Krzanich added that, had a 1971 Volkswagen Beetle improved at the same rate as Intel microchips, “You would be able to go with that car 300,000 miles per hour. You would get two million miles per gallon of gas, and all that for the mere cost of 4 cents!” Moreover, that car would be the size of an ant. Your smartphone contains significantly more computing power than the machines that sent the Apollo astronauts to the Moon. Such hardware advances have provided the raw brainpower not only for the AI explosion but also for the digitalization of modern society.

The second major factor driving the AI revolution has been remarkable advancement in ML. The goal of ML is to train a computer to perform some cognitive task, for example, recognizing objects within images or interpreting spoken language. One common approach called supervised learning trains the digital system, often a neural network—a pliant digital lattice designed loosely to emulate the type of connectivity found in human brains—with a very large collection of samples of the items to be recognized, appropriately labeled. The system is then tested on unlabeled samples. A correct identification causes certain pathways through the network to be reinforced and others reduced. Over time recognition improves as the system literally rewires itself.

A digital translator, for example, might be trained on a large collection of English documents paired with good Chinese equivalents. It is then set loose on English text without corresponding translations. A human or automated trainer scores each translation and this feedback prompts strengthening of successful pathways and degrading of unsuccessful ones. It is this approach that led to remarkable recent advances in Google Translate; after being converted to an ML approach, the system “demonstrated overnight improvements roughly equal to the total gains the old one had accrued over its entire [10 year] lifetime” (Lewis‐Kraus ).

The holy grail of ML, however, is unsupervised learning, in which the network extracts meaningful patterns without being told explicitly what to search for. In principle, a digital system could learn English to Chinese translation without access to paired documents. According to a recent article in Science,

That's possible because languages have strong similarities in the ways words cluster around one another. The words for table and chair, for example, are frequently used together in all languages. So if a computer maps out these co‐occurrences like a giant road atlas with words for cities, the maps for different languages will resemble each other, just with different names. A computer can then figure out the best way to overlay one atlas on another. Voilà! You have a bilingual dictionary. (Hutson )

The developing system is stressed by translating sentences from language A to language B and back again. If the result is identical to the original sentence, then the pathway is reinforced; otherwise, it is weakened. Recent advances using this approach promise great benefits for translations among languages, including obscure and endangered languages for which there are not vast corpora of documents with paired translations.

ML using neural nets is not a new technique; the earliest neural models date from the 1940s. But they fell out of favor for decades as other techniques exhibited faster progress. The main problem: simple neural nets can only do simple things. Real progress required the advances in hardware presaged by Moore's Law, the accumulation of vast corpora of training data, and development of “multilayered” architectures. Layering allows the system to search for patterns, then patterns of patterns, and so on. A system to understand text, for example, might comprise a lowest layer to recognize individual words, which are passed to a second layer that supplies a likely meaning gleaned from the context (adjoining words), which are passed to a third layer that assembles them into recognizable phrases, then into recognizable sentences, and so on.

The accelerating progress in intelligent computational systems along with the growing facility of robots on a wide variety of tasks raises an age‐old question: How will these emerging technologies impact the workforce? Will they lead to widespread unemployment as many human skills become redundant? Or will they create new employment opportunities that offset their disruptive effects? Such questions have arisen at least since the first industrial revolution when English textile workers battled the introduction of mechanized looms under the banner of the folkloric Ned Ludd. (Some scholars describe the rise of AI as a fourth industrial revolution [Schwab ] or second machine age [Brynjolfsson and McAfee ].) Studies, however, consistently have shown that technological innovation, while disruptive to certain professions, historically has produced net increases not only in productivity, but also in employment.

Economists steeped in the history of productivity improvement tend to point out that it is always bad in the messy middle of a transformation, but everything works for the better in the end, as productivity gains give rise to investments in new ventures that provide jobs that were never imagined before. At least historically that has always been true. But others insist that this time, history will not rhyme (Davenport and Kirby , 227).

Historically, it has been blue‐collar workers who were supplanted by automated systems. That trend continues. Tasks like ordering a hamburger, buying groceries, and shipping goods require drastically fewer human workers than previously. Self‐driving cars are coming; but so are self‐driving trucks (Freedman ) and even crewless cargo ships (Levander ). Robots now lay bricks, operate mining equipment, and make pizzas. Entire classes of jobs have been virtually eliminated: switchboard operators, filing clerks, travel agents, and assembly‐line workers. The MIT Technology Review reports (Rotman ) that 83 percent of current jobs that pay less than $20 an hour are under threat from automation. According to a study commissioned by President Obama, the jobs “threatened by automation are highly concentrated among lower‐paid, lower‐skilled, and less educated workers” (White House , 2).

Increasingly, however, white‐collar and professional jobs are seeing incursions from automation. Physicians increasingly consult computer‐based diagnostic assistants such as DXplain and VisualDx to confirm their diagnoses or suggest alternatives (Maron ). A legal assistance tool called ROSS, leveraging the brain power of IBM's artificially intelligent supercomputer Watson, can perform in seconds legal research that once took a skilled paralegal hours of work. Each quarter, Associated Press earnings reports for over 3,700 companies go “out to the wire” without any human intervention; the stories are written entirely by software, represent a twelve‐fold increase over manual output, and contain fewer errors than the human‐produced stories from years past (Miller ). Up to 70 percent of Wall Street trades are made by computerized high‐frequency trading systems (Barrat , 94). The professions of medicine, law, journalism, and finance likely will never be the same. But, surely, clergy are immune from the usurpations of automation. Or are they?

In the following two sections, we will outline ways in which technology may impact ministry both in the near term and in some plausible futures. First, we will outline ways in which current technologies are changing the religious landscape. Then, in the following section, we will speculate about how future AI systems might usurp some of the traditional functions of clergy, specifically preaching and pastoral care.

ONLINE SPIRITUALITY

Recall that by online spirituality we denote the delivery of religious products and services via digital media outside of traditional congregational settings. This is more than just using technology in church. Tech‐savvy pastors have long supplemented their liturgy, congregational communication, evangelization, and church management with high‐tech, particularly in mega‐churches that can afford the hardware, software, and support staff needed to run a tech‐heavy organization. Some mega‐churches even use telepresence to allow the pastor to preach simultaneously at multiple campuses (Blake ).

Many churches, both large and small, stream worship services online or make them available to be watched anytime on YouTube. Increasingly, accessible web development tools, blogging sites, and new media make it easy even for small church pastors to have a significant digital presence. This facilitates access to their message, allowing congregants, including the elderly or infirm, remote access to congregational resources. It also allows potential congregants to sample the offerings of the local church prior to a visit. But offering online access to these products and services means they become available also to a broader audience, many of whom may have no interest in affiliating with or supporting the local church.

Church leaders recognize the value of the enhanced visibility that digital access affords. Pope Benedict XVI in his 2010 message for World Communication Day challenged ministers “to proclaim the Gospel by employing the latest generations of audiovisual resources (images, video, animated features, blogs, websites) [to] open up new vistas for dialogue, evangelization and catechesis” (Pope Benedict XVI ). Nearly every denomination now has a web presence, often providing extensive doctrinal, liturgical, and counseling resources. Mormon.org, for example, provides 24/7 access to Latter‐Day Saints missionaries on call; it is reported that this multiplies the conversion rate of a missionary by sevenfold over door‐to‐door evangelization (Bosker ). Religious leaders increasingly communicate via social media; the Pope has more than 19 million Twitter followers, the Dalai Lama more than 10 million. Spirituality‐related courses and web‐based retreats are available from hundreds of providers.

A 2013 study (Buie and Blythe ) found over six thousand apps on the iTunes App Store related to spirituality and religion, including selections for education, prayer exchange, comparative religion, guidance for daily living, and teachings and sermons from specific religious leaders. A Pew Research study (Pew Research Center ) reported that around 20 percent of American adults used social media to share their religious faith, “about the same percentage that tune in to religious talk radio, watch religious TV programs or listen to Christian rock music.”

Electronic media provide access to the products of religion, but without any direct contact with clergy. It is possible to sample one hundred different sermons every day on YouTube without exhausting the ever‐replenished pool. Many clergy also publish blogs to spread their message; a Google search for “clergy blog” returns almost 900,000 results. The facile online availability of religious goods and services almost certainly exacerbates the ongoing erosion of traditional religious observance, specifically attendance at brick and mortar houses of worship.

As religious products become ever more readily accessible, they are also increasingly becoming mere commodities. Why bother reading a blog or listening to a sermon from your local pastor when you have instantaneous access to some of the most charismatic religious leaders in the world? This trend is already evident as TV preachers such as Kenneth Copeland, T. D. Jakes, and Joel Osteen cultivate extensive “virtual” congregations, in addition to the mega‐churches they pastor. Such factors could drive the religious landscape in the direction of what Duke University economist Philip Cook and others have called a “winner‐take‐all market,” where a handful of superstar performers garner a larger and larger proportion of the available audience, and reap the lion's share of the rewards. This pattern, long common in entertainment and sports, is becoming increasingly prevalent in other fields—law, journalism, consulting, investment banking, management, and fashion (Frank ). It could become the new norm in the religious “marketplace” as well.

As most religious services become available remotely, the demand for local clergy shrinks. It is easy to imagine a plausible future in which a pool of clergy for hire fulfills residual demand for those occasions when a local clergyperson is required, say, to officiate at a wedding or funeral. Alternatively, a quick visit to the Universal Life Church (ULC) website can result in an online ordination. With more than twenty million ordained “ministers,” ULC “has become one of Earth's largest and most active religious organizations” (Universal Life Church ). For the Susskinds and others, these trends suggest that religion and spirituality are trending in the do‐it‐yourself direction, potentially significantly reducing the need for local professional clergy.

Clearly, many individuals still will desire the fellowship of a local church community and an ongoing relationship with a spiritual mentor. Moreover, some studies have found that online religious activity generally serves to supplement rather than supplant offline church activity (Campbell and Garner , 66–67). Nevertheless, the “nones” and the “spiritual but not religious” are among the fastest growing groups in the American religious landscape. Is this any surprise when the products of religion are so readily available without the commitment, inconvenience, or expense of joining a local congregation on Sunday morning?

DIGITAL CLERGY

In this section, we venture beyond what is currently happening in the online realm to speculate about the following question: Could automated systems ever supplement or supplant clergy in such core functions as writing/delivering sermons or providing pastoral care? Current technological trends in established professions such as law, medicine, journalism, and finance suggest that the answer is yes. Whether or not it is a theologically or practically sound idea is a separate question. We will return to it in the section entitled “Theological Issues.” For the remainder of this section, we consider what assumptions underlie our belief that digital clergy are likely to emerge in the coming decades, and then speculate about what that might look like, specifically in the areas of preaching and pastoral care.

THEOLOGICAL ISSUES

Some profound theological issues arise from the very prospect of advanced AI. To name just a few:

• What is the ontological status of a truly intelligent automated system, assuming such an entity is possible? Could such a system have a soul? Would it have rights?

• What would the existence of such a system say about the unique status of human beings created, according to Judeo‐Christian theology, in the image of God?

• What constraints does morality impose on a scientific community racing to develop technologies that at the very least will have profound social and economic impacts and, according to some observers, could threaten the very existence of humanity?

• What are the ethical implications of automation reshaping the workplace with the concomitant effects on human dignity and autonomy?

No doubt, these issues will occupy the scientific and theological communities for decades to come. Some have been addressed elsewhere. See, for example, Herzfeld () and Campbell and Garner ().

Here, we will focus narrowly on theological issues raised by the prospect of automated systems undertaking functions currently performed by ordained clergy. Our goal is not to hash out these issues that arise from digital ministry, but merely to raise them. If and when clergybots appear, theologians undoubtedly will engage these issues more fully.

Can a digital system minister? Assume that an automated system could provide precisely the same empathetic emotional and spiritual support as your current clergyperson. Is that even possible? Would it constitute ministry in the sense that we currently understand ministry, or in some new sense?

Can a digital system administer sacraments? Faith traditions that recognize sacraments may also specify who is allowed to confer the sacrament. For example, the Catechism of the Catholic Church specifies that “any person, even someone not baptized, can baptize, if he has the required intention.” Such a baptism is always valid, but sometimes illicit. Suppose the baptism were conferred by a robot, pronouncing the appropriate trinitarian formula. Would this satisfy the requirement of “any person”? What about the “required intention”?

Where does faith enter in? Could one legitimately ascribe to a digital system faith in any meaningful sense? Would it matter? If a sermon is spiritually insightful and moving, would it matter that the author does not/could not believe a word of it? As with the AI palliative care agent described above, if a system functions equally competently within Buddhist, Hindu, and Christian contexts, could it have any religious credibility or authority?

Could a digital system be ordained? Traditionally, religious authority has been established by a process of discernment of one's spiritual calling or divine appointment, or by undergoing an extensive period of training. What does it imply for religious authority if spiritual formation is replaced by uploading? What does ordination imply about the status of the ordinand? Is there a meaningful sense in which a digital system can be said to subscribe to the tenets of one or another denominational orthodoxy?

What about relationships? Many theologians within the Christian tradition (e.g., Karl Barth, Emil Brunner) aver that humans reflect the imago Dei precisely in their capacity for social relationships with God and other people. Can humans form genuine relationships with machines allowing for the type of intimate engagement required for ministry to occur?

These are just a few of the issues that arise from the prospect of digital clergy. There is a danger that even considering such questions might be seen as devaluing the profession of (human) clergy. However, as the digital systems become ever more capable, and human members of society become ever more comfortable interacting with automation, there may come a time when such considerations will become necessary.

CONCLUSIONS

Despite decades of effort, the field of AI research is still far from achieving artificial general intelligence (AGI). Current AI systems have been likened to “autistic savants,” extraordinarily capable within a very limited range of tasks but effectively useless outside that range. But that is not from lack of trying. We saw above the consensus of AI experts that HLMI likely will be attained by the end of the century, if not well before. While AI experts have been notoriously overly optimistic in the past about the trajectory of AI progress, recent efforts have led to some breathtaking breakthroughs that lend credence to these predictions.

Existing systems show that intellectual and creative tasks we currently reserve for human beings can be automated. The fields of medicine, law, journalism, and finance are changing rapidly to adapt to the new technologies. Could the profession of clergy be next? Systems described above—Project Debater, the palliative care chatbot, and others—suggest that sermon writing, pastoral care, and certain other tasks currently performed by clergy could be automated were there sufficient incentives to do so. Focus in AI research is driven by intellectual curiosity on the part of researchers, but also increasingly by strong economic factors. Time will tell if clergy tasks will provide a tempting target for researchers, or if efforts in related fields may reach sufficient maturity to allow them to be repurposed to automate clerical tasks.

In the near term, the artifacts of online spirituality likely will become increasingly available, potentially exacerbating the ongoing decline of attendance at brick and mortar houses of worship. Although clergy are not currently in the crosshairs of automation, there is no guarantee that that situation will continue. As has happened in other professions, it seems likely that increasingly capable AI systems will encroach inexorably on the tasks and prerogatives that define us as human. Theologians will have to confront the issues raised by these developments. All that is certain is that progress in AI will continue apace. How the results will reshape the religious landscape is anyone's guess.